A recent increase in the ease and availability of alternative rides for intoxicated passengers partially explains the steep decrease in alcohol-related collisions in New York City since 2011. This study examines the specific case of Uber’s car service that was launched in New York City in May 2011, a unique example of a sudden increase in cab availability for intoxicated passengers.
The study draws on a dataset of all New York State alcohol-related collisions maintained by the New York State Department of Motor Vehicles. The full estimation sample contains 4,526 observations of monthly data in 62 New York State counties, five of which are New York City boroughs, from January 2007 through July 2013. This period includes Uber's entrance into the New York market in May 2011, allows for a substantial pre-intervention period, and omits two potentially confounding entrances in the New York City transportation market that could influence the alcohol-related collision rate.
A difference-in-differences estimation was produced using four different control group specifications based on different criteria exogenous to the intervention. Inferences were based on the variation in Uber access across New York State counties over time and the careful choice of New York State counties that provide an appropriate control group for New York City's drunk-driving behavior.
Econometric analyses show that each of the New York City boroughs that experienced significant Uber service coverage (Manhattan, the Bronx, Brooklyn, and Queens) experienced a 25 to 35 percent decrease in its alcohol-related collision rate using difference-indifferences estimation and standard errors clustered at the county level. The most conservative estimate of this intent-to-treat (ITT) effect implies a decrease of approximately 43 crashes per month across the New York City boroughs with Uber coverage, based on pre-period collision rates and population.
Apportioning the average effect across the four treated counties by their population of vehicles implies:
- A crash count decrease in Queens of 16 to 22 crashes per month, a 33 to 45 percent decrease from Queens' 2010 average monthly alcohol-related crash count.
- A crash count decrease in Manhattan of 5 to 7 crashes per month, a 16 to 22 percent decrease based on Manhattan's 2010 average alcohol-related crash count.
(Our website has many links to other organizations. While we offer these electronic linkages for your convenience in accessing transportation-related information, please be aware that when you exit our website, the privacy and accessibility policies stated on our website may not be the same as that on other websites.)