Uber economic study: Uber serves underserved neighborhoods in Chicago as well as the Loop. Does taxi?
The Uber app was created to make sure that people could get rides when and where they need, no matter what. We already know that 4 in 10 rides in Chicago start or end in underserved neighborhoods and we have heard story after story from riders around the world who have never been able to get a taxi to come to their neighborhood, and who now rely on Uber for reliable transportation.
So we decided to take it a step further and investigate the quality of service in those communities. This study explores what, if any, relationship exists between a neighborhood’s economic well-being and the level of service that Uber’s network provides in that neighborhood. We studied two related questions: 1) does a neighborhood’s median income predict the expected time before a driver appears after a request is made? and 2) does a neighborhood’s median income predict the likelihood that a request will go completely unfulfilled?
The results? Median income in a neighborhood has no meaningful relationship to Uber’s level of service in that neighborhood, including wait times and fulfillment rates.
Methodology: The wonky stuff
In order to investigate the relationship between local income and Uber’s level of service, we analyzed a data set that included:
median income by neighborhood
neighborhood in which each trip originated
whether or not each trip was ultimately fulfilled
the time between request and pickup for each completed trip
The relationship between a neighborhood’s income and the waiting time was determined by running an ordinary least squares regression of the form wait = a + b*income. This regression was run on a large number of trips in 2014 in Chicago. The relationship between a neighborhood’s income and the probability of a ride’s being fulfilled was determined with a similar specification, prob_unfulfilled = a + b*income. This second regression was run on a single observation for each neighborhood of the percentage of trips that are not fulfilled as a percentage of all trip requests.
Results: The numbers
Median income is related to expected wait time by the formula
wait (in seconds) = 510 – .003 * median income (in dollars)
In other words, a $1,000 increase in a neighborhood’s median income is associated with 3 fewer seconds in expected wait time. The data and regression line are shown graphically below.
Median income is related to the probability of a ride going unfulfilled by the formula
prob_unfulfilled = .3 – .000003 * median income (in dollars)
In other words, an increase in median income of $1,000 is associated with an increase in the probability of fulfillment of .003 (ie, .3 percentage points). The data and regression line are shown graphically below.