Getting Home Safe: The Effect of Ridesharing on Drunk Driving — A Regression Discontinuity Approach

Abstract

In the last half-decade, ridesharing services have emerged as an ultra-popular means of transportation in urban and suburban communities. With the supreme convenience of requesting a ride directly from a smartphone and removing the hassle of payment, ridesharing apps have quickly supplanted taxis as the primary alternative to personal vehicles and public transportation. Despite the popularity of these services, however, their public welfare effects are still largely unknown. This paper seeks to evaluate the effects of ridesharing as an alternative to drunk driving using traffic data from Austin, Texas between 2015 and 2017. After a drawn-out public dispute over legislation to implement more rigorous background checks for drivers, the two largest ridesharing companies, Uber and Lyft, suspended their services in Austin on May 9, 2017. Just over a year later, they returned to the city and have continued to operate since. This yearlong window provides an ideal natural experiment to analyze their effect on drunk driving. Using a regression discontinuity analysis, I find that ridesharing has high potential to reduce incidence of dangerous and drunk driving. This effect did not translate into fewer injuries or deaths, however the data shows significant drop-offs in alcohol-related crashes after the companies’ return to Austin, suggesting potentially large positive externalities associated with their services.