The Harvard Data Science Initiative (HDSI) Faculty Special Projects Fund annually allocates a total of $50,000 to support one-time data science opportunities throughout the year. In June 2021, Harvard Assistant Professor of Economics and HDSI Faculty Affiliate, Gabriel Kreindler, was awarded a $5,000 grant for his research project, “A Toolkit for Precise Geographic Data on Urban Roads: Application to Measuring the Impact of Automated Cameras on Speeding and Road Crashes.” Kreindler specializes in studying urban transportation issues in developing countries, such as the impact of traffic congestion management policies in major cities.
For this HDSI-funded project, Kreindler collaborated with his Research Assistant, Brendan Chapuis (‘22), to create a toolkit for precisely map-matching geospatial data to a geographic base map in the context of studying the impact of installing automated cameras on speeding and road crashes in São Paulo, Brazil.
Their goal was to develop the code and best practices to match geographic events (accidents, camera locations) to the road-segment level base map (based on Open Street Maps). The highly local behavioral trends of subjects under study required a high level of geographic precision, down to the level of individual road segments.
During the summer and fall of 2021, Chapuis, who was a Harvard undergraduate studying economics at the start of this project, used the uniquely rich data accessed through the universe of camera installs and geocoded accidents between 2012 to 2014 and 2018 in São Paulo to develop the required code and begin the analysis for their research.
“Brendan did a phenomenal job in this project, creating the code base, as well as publishing it in a public repository on GitHub, and writing a Medium article that explains the process and serves as a ‘walk-through’ for similar applications,” said Kreindler. “We hope that these outputs may be useful for other projects that tackle similar problems as we did.”
Kreindler and Chapuis believe that their toolkit will aid in approaching other data science applications related to urban road transportation and continue to work on this research project using the code developed with the financial support of the HDSI.