Periodic Reporting for period 1 - ride-hailing (Estimating the effect of ride-hailing on public transportation)
Reporting period: 2021-07-01 to 2022-06-30
The explosive growth of ride-hailing around the world has sparked an important debate about the repercussions this new mobility option is having on cities. Urban planners and policymakers around the world have grappled with how to regulate ride-hailing companies in their jurisdictions. Indeed, several countries have banned ride-hailing while others have heavily regulated it. Of particular interest to urban planners and policy makers is whether ride-hailing technologies increase or decrease public transit ridership. Understanding the degree of complementarity or substitutability between ride-hailing and public transit is important for at least four reasons. First, reductions in transit ridership can potentially generate major budgetary shortfalls for transit authorities. Second, reductions in transit ridership likely have social welfare costs because transit ridership is inefficiently too low. Third, reductions in transit ridership likely increase congestion and pollution. Fourth, changes in transportation technologies, such as steam railways, the automobile, and limited-access highways, have repeatedly reshaped urban spatial structure. However, there remains great uncertainty about how ride-hailing and public transportation are impacting each other.
This project addresses this problem by estimating how the rise of ride-hailing is changing cities and the lives of those living there. I have two specific research objectives:
- Estimate how ride-hailing affects public transit. This question is contentious, with valid theoretical arguments on either side. I will do so by using Uber’s confidential trip data to estimate the effect of new subway stations on Uber ridership. While this is estimating how transit affects Uber, rather than how Uber affects transit, it provides a new approach to answer the question of whether Uber and public transit are complements or substitutes.
- Discover the mechanisms by which public transit and ride-hailing affect each other. This is especially useful because it will help with designing policies to encourage public transit ridership and help with the design of public transit networks.
Furthermore, this project also provides preliminary insight into the impact of another technological innovation within urban transportation: autonomous vehicles. While there is much speculation about how autonomous vehicles may change cities, no empirical estimates exist to date because the technology is so new. However, since the change from taxi to ride-hailing is similar to the change from ride-hailing to autonomous vehicles, in that autonomous vehicles will make transportation more convenient, accessible, and affordable, then these results will inform us about this next technological innovation’s effect on cities and will help cities prepare for the change.
This project addresses this problem by estimating how the rise of ride-hailing is changing cities and the lives of those living there. I have two specific research objectives:
- Estimate how ride-hailing affects public transit. This question is contentious, with valid theoretical arguments on either side. I will do so by using Uber’s confidential trip data to estimate the effect of new subway stations on Uber ridership. While this is estimating how transit affects Uber, rather than how Uber affects transit, it provides a new approach to answer the question of whether Uber and public transit are complements or substitutes.
- Discover the mechanisms by which public transit and ride-hailing affect each other. This is especially useful because it will help with designing policies to encourage public transit ridership and help with the design of public transit networks.
Furthermore, this project also provides preliminary insight into the impact of another technological innovation within urban transportation: autonomous vehicles. While there is much speculation about how autonomous vehicles may change cities, no empirical estimates exist to date because the technology is so new. However, since the change from taxi to ride-hailing is similar to the change from ride-hailing to autonomous vehicles, in that autonomous vehicles will make transportation more convenient, accessible, and affordable, then these results will inform us about this next technological innovation’s effect on cities and will help cities prepare for the change.
The work performed from the beginning of the project to the end of the period covered by this report includes:
- Preparing data on new train stations
- Identifying all the new train station openings in cities Uber operates
- Classifying all the new train station openings by the type of service provided
- Calculating the population density, weather, and nighttime luminosity near transit stations using pre-existing datasets
- Preparing aggregate data on Uber trips. We use Uber's trip database to calculate the number of trips starting or stopping within a given geography in a month. We use two different geographies.
- The first geography is distance bands. We use Uber's trip database to calculate the number of trips starting or stopping within a given distance band from the transit station in a month, for example, all trips within 100--200 meters. We do this for 12 different 100 m bands from 0--100 m up to 1,100--1,200 m.
- The second geography is hexagons. These hexagons are defined using the H3 hexagonal hierarchical geospatial indexing system, designed by Uber. We have data on all hexagons whose centroids are within 1,200 m of a transit station.
- Preparing data on Uber trips for subgroups. In order to identify the mechanisms by which ride-hailing and transit impact each other, we identify subgroups that are impacted by one mechanism, but not others, and test for the impact of the new transit station opening on them.
- Analysis
- Estimate the aggregate impact of a new transit station opening on Uber rides at all distances up to 1200 meters
- Estimate whether the impact of a new transit station opening on Uber rides differs by city, continent, time-of-day, weather conditions, population density
- Estimate the impact of a new transit station opening on Uber trip length
- Conduct robustness tests
- Conduct a placebo test where we randomly assign placebo dates for the opening of the new transit station
- Use alternative statistical estimators, including pseudo-poisson maximum likelihood
- Write and post a working paper containing current results
- Get feedback on current results from researchers at VU, LSE, and Copenhagen Business School (and others)
- Preparing data on new train stations
- Identifying all the new train station openings in cities Uber operates
- Classifying all the new train station openings by the type of service provided
- Calculating the population density, weather, and nighttime luminosity near transit stations using pre-existing datasets
- Preparing aggregate data on Uber trips. We use Uber's trip database to calculate the number of trips starting or stopping within a given geography in a month. We use two different geographies.
- The first geography is distance bands. We use Uber's trip database to calculate the number of trips starting or stopping within a given distance band from the transit station in a month, for example, all trips within 100--200 meters. We do this for 12 different 100 m bands from 0--100 m up to 1,100--1,200 m.
- The second geography is hexagons. These hexagons are defined using the H3 hexagonal hierarchical geospatial indexing system, designed by Uber. We have data on all hexagons whose centroids are within 1,200 m of a transit station.
- Preparing data on Uber trips for subgroups. In order to identify the mechanisms by which ride-hailing and transit impact each other, we identify subgroups that are impacted by one mechanism, but not others, and test for the impact of the new transit station opening on them.
- Analysis
- Estimate the aggregate impact of a new transit station opening on Uber rides at all distances up to 1200 meters
- Estimate whether the impact of a new transit station opening on Uber rides differs by city, continent, time-of-day, weather conditions, population density
- Estimate the impact of a new transit station opening on Uber trip length
- Conduct robustness tests
- Conduct a placebo test where we randomly assign placebo dates for the opening of the new transit station
- Use alternative statistical estimators, including pseudo-poisson maximum likelihood
- Write and post a working paper containing current results
- Get feedback on current results from researchers at VU, LSE, and Copenhagen Business School (and others)
This proposal’s approach expands beyond the state-of-the-art by using a better-identified estimation strategy using microdata to estimate the how Uber affects public transit for more than 80 cities around the world with current UberX service that have also experienced recent subway expansions. Since the expansion of subway systems is usually a long process subject to delays and planned long before Uber became a feasible alternative, this offers an ideal setting for studying the interactions between Uber and public transit. The second key innovation is the use of individual-level longitudinal data on travel behavior (instead of metropolitan area aggregates), which allows me to address the mechanisms by which Uber affects public transit.
Our current findings show that when a new train station opens, Uber ridership strongly increases very close to the station. This shows that Uber is often used with trips by train to help with the first or last portion of the journey, and so helps with the first/last mile problem. We continue to investigate the other mechanisms by which Uber and transit impact each other and the magnitude of the first/last mile effect relative to these other mechanisms.
Our current findings that the relationship between ride-hailing and public transportation are stable across cities and countries is also valuable, as it suggests that studies of a single city or country have external validity and apply to other cities and countries.
This research has broader social impacts as well. This research will improve our understanding of how ride-hailing and public transit affect each other. This will lead to better-informed regulation of ride-hailing, better-designed policies to help ride-hailing complement public transit, and better public transit investment decisions. Since, from the passenger’s perspective, ride-hailing is very similar to an autonomous vehicle, this research also will help cities and transit agencies prepare for autonomous vehicles. Well-informed regulation and policies that lead to more shared rides, either by increasing public transportation ridership or by shared rides in ride-hailing, will reduce personal automobile traffic, reducing traffic congestion and pollution. By reducing congestion and ensuring ride-hailing complements transit, we increase mobility within cities, which increases the quality-of-life within cities.
Our current findings show that when a new train station opens, Uber ridership strongly increases very close to the station. This shows that Uber is often used with trips by train to help with the first or last portion of the journey, and so helps with the first/last mile problem. We continue to investigate the other mechanisms by which Uber and transit impact each other and the magnitude of the first/last mile effect relative to these other mechanisms.
Our current findings that the relationship between ride-hailing and public transportation are stable across cities and countries is also valuable, as it suggests that studies of a single city or country have external validity and apply to other cities and countries.
This research has broader social impacts as well. This research will improve our understanding of how ride-hailing and public transit affect each other. This will lead to better-informed regulation of ride-hailing, better-designed policies to help ride-hailing complement public transit, and better public transit investment decisions. Since, from the passenger’s perspective, ride-hailing is very similar to an autonomous vehicle, this research also will help cities and transit agencies prepare for autonomous vehicles. Well-informed regulation and policies that lead to more shared rides, either by increasing public transportation ridership or by shared rides in ride-hailing, will reduce personal automobile traffic, reducing traffic congestion and pollution. By reducing congestion and ensuring ride-hailing complements transit, we increase mobility within cities, which increases the quality-of-life within cities.