| By Allison Crow and Edward J. Klock-McCook, RMI
The extent to which TNCs add or subtract from traffic is a function of three things: 1) TNCs repositioning (deadheading) and circling while looking for a fare (trolling), 2) whether passengers would have used active modes (walking/biking) instead, or 3) TNCs loading and unloading passengers in the middle of thru lanes. The effects of the latter two—mode-shift and loading and unloading—will be tackled in future work.
The VMT Efficiency Metric
Cities need a reliable metric that determines how TNCs are impacting vehicle miles traveled (VMT). VMT efficiency is a means of doing just that. It is the ratio of the mileage a trip would take a person to drive to his or her destination using a personally owned vehicle (including distance traveled to find parking) to the total mileage it takes the TNC to drive a person to the same destination.
If the efficiency is greater than one, then TNCs are more efficient than personally owned vehicles. The average VMT efficiency of TNCs for a city helps quantify the impact TNCs are having on traffic.
Lyft collaborated with RMI and shared trip data from November 1, 2016, to October 31, 2017, for San Francisco, Chicago, and New York (Figure 1). As some of Lyft’s more well-developed markets, these urban areas span a variety of geographies and climates and include the dense urban core as well as the less dense perimeter. These data allowed RMI to begin to answer questions like: do TNCs add or subtract miles relative to typical personally owned vehicles? What factors influence this and how can they be measured? This analysis unveils new insights into the way Lyft functions, identifies where questions remain unanswered, and reveal insights for ways to answer those questions.
Tracking Mileage Driven
Lyft tracks ride data through its app in three driver operating modes, as shown in Figure 2. Period 1 is mileage driven while a driver has his or her app on but has not yet been matched with a ride. When a ride is requested and a passenger is paired with a driver, the driver enters period 2—the distance the driver must travel to pick up his or her passenger(s). Period 3 is the passenger ride to the destination. The total mileage per trip for each metropolitan area and the breakdown of that mileage for each period is shown in Figure 3.
A closer look at each of these periods gives a better understanding of the impact TNCs have on cities and possible ways to improve the TNC efficiency in those cities. Breaking the data down by period, especially period 1, also shows where there are gaps in information that lead to a conservative calculation of TNC efficiency.
Period 1: TNC app on but no ride active
Lyft’s period 1 is the mileage drivers travel with their app activated without an active ride match. Miles of this type could come from a variety of sources, some of which decrease efficiency while others do not. Miles that are the result of drivers deadheading or trolling add VMT to the system. In contrast, since TNC drivers are independent contractors, some mileage they travel is for personal reasons, not adding additional VMT to the system. For example drivers may have the app on during their commute in case an opportunity arises to give a ride along the way. In addition, drivers often run multiple TNC apps at the same time to pick up more rides. SherpaShare estimates that 75 percent of drivers drive for two or more TNCs meaning that some of the mileage that shows up in Lyft’s period 1 data may overlap with another TNC ride. At present, these different sources of mileage cannot be disaggregated, meaning this portion of TNC impact, treated conservatively here, contains the greatest uncertainty.
Period 2: Travel to pick up
When a passenger summons a ride on Lyft, an algorithm matches that passenger or passengers to a driver in the area and the driver is given directions to the pick-up point. Period 2 is the mileage the driver must travel to get to that location. These miles are a function of TNC market penetration because as ridership and the number of drivers increases, a driver is more likely to be closer to a ride request. Anecdotally, this is already apparent today when observing longer wait times in smaller markets compared to those in major markets.
Period 3: The ride
The final data set is period 3, the passenger’s ride to his or her destination. Though seemingly straightforward, the trip length as reported by Lyft is affected by the inclusion of Lyft Line rides. Lyft Line is available in most markets and enables parties of one or two to save money by adding another party to their trip when the trip routes overlap. It is important to understand that the period 3 mileage for a Lyft Line trip is recorded from the beginning of the first Lyft Line passenger until the end of the last passenger. The impact of this nuance occurs on a spectrum. At one end of the spectrum, if the passengers overlap for only a couple blocks out of several miles, the efficiency derived from the data is fairly accurate. On the other hand, if the passengers overlap for all but a few blocks, then the efficiency is nearly double what the data would suggest because that trip, which is really two rides given at the same time, is reported as only one. Therefore, the results presented above are on the conservative end of that spectrum because the mileage associated with multiple-person trips are aggregated into one reported trip.
Lyft Line rides improve the load factor and thus improve the VMT efficiency of TNCs. Not all ride requests are for Lyft Line and not all Lyft Line requests are matched with other riders. Lyft records the number of regular Lyft rides versus Lyft Line rides requested, as well as the number of Lyft Line rides matched—in which another party is added to the ride. Lyft Line requests in the three cities are approximately one-third of total Lyft rides. This breakdown gives insight into consumer choices around rideshare, however the total number of passengers in each ride and how long their travel overlaps is still missing.
More complete data is needed to understand actual passenger miles traveled as compared with simply vehicle miles traveled. After all, movement of people is the priority of the city and the most efficient city will move everyone, everywhere, with the minimum number of vehicle miles traveled.
Estimating VMT Efficiency
From a simple analysis using the well-measured data shown in Figure 1, Lyft is about 20 percent more efficient at moving people through a city than a personal car, however not all factors contributing to the TNC VMT efficiency are well-measured, making this estimate conservative. To try to further illuminate the actual efficiency, Figure 4 uses San Francisco as an example to show possible efficiencies due to not having to look for parking, sharing rides, data transparency, and system optimization. As outlined above, the challenge is that many of these factors have a degree of uncertainty and some factors are not yet even possible to individually quantify. It’s clear that a TNC trip does not involve looking for parking, but quantifying the avoided miles is difficult. Data clarity will help disaggregate the period 1 miles discussed above. Future system improvements include things like pick-up and drop-off zones, market penetration, and even autonomous vehicles.
Figure 4 shows that TNCs have additional efficiencies that, when factored in, could raise the VMT efficiency of Lyft to 60 percent better than a personally owned vehicle. This result suggests two complementary actions for TNCs and cities going forward to realize maximum efficiency. The first is to improve the quality, availability, and clarity of vehicle mileage data. The second is to take additional actions toward improving the system as a whole. Examples of this include short-term parking in high-demand areas, seamless integration with public transit, and dedicated pick-up and drop-off areas.
This is a moment of opportunity for the TNC industry and the transportation sector to help grow a new service model in the best possible way. In order to seize this opportunity, the first step is to know what we are working with. Using the VMT efficiency calculation, we have identified that, compared with personally owned vehicles, TNCs have fewer vehicle miles traveled per trip. As a result, cities could benefit by leveraging TNCs for those unwilling or unable to use transit or active modes.
TNC efficiency is the first step in RMI’s effort to create a conversation around the optimization of TNCs’ impact on congestion, convenience, mobility cost, and carbon emissions. Our plan is to engage with TNCs, cities, and other stakeholders to develop new data sources, understand where this metric is valuable, and fill in the gaps to provide the answers cities need to maximize the benefits of the transportation service economy.
Image courtesy of iStock.
