Analysis of Travel Choices and Scenarios for Sharing Rides
Final Report

Executive Summary

Automobile occupancy has important implications for the efficiency of highway operations. Increasing automobile occupancy by sharing rides may help reduce overall vehicle miles traveled (VMT), and thereby alleviate congestion, curtail vehicle emissions, and support economic growth.

Two approaches to facilitate higher automobile occupancy have emerged in practice and from pilot studies. They are:

• Shared options in ridehailing applications. Ridehailing applications have significantly impacted the transportation landscape by providing an alternative to personal vehicles, and in some cases, an alternative to public transportation. These services, offered by transportation network companies (TNC), such as Uber™, Lyft™, and Via™, are available through mobility apps. These mobility apps allow users to select private rides, such as UberX™ or a standard Lyft™, or shared rides, such as UberPool™, Shared Lyft™, or Via™. The apps provide an upfront estimate on price and time for all options presented, enabling users to make calculated decisions about their travel behavior.

• Ridesharing using app-based incentive tools. Emerging app-based ridesharing models continue to augment traditional carpooling by providing features that allow dynamic ridematching (with matches created in real-time and through automated mechanisms including through the establishment of trusted networks), incentive programs, occupancy verification and other options that allow for more flexible carpool formation, higher user satisfaction and sustained use of these services.

This study seeks to understand behavior and the impacts of price and time tradeoffs for both the aforementioned models of sharing private rides. By analyzing the tradeoffs users make when presented the option to share a ride, typically with incentives to do so, this study may be able to discern impacts of various choice and incentive structures.

While transportation researchers have analyzed ridehailing behavior before, there is limited research describing the effect of price and time on a rider's choice between private party and shared ridehailing. This Federal Highway Administration (FHWA) study used data on revealed preferences for private party and shared ridehailing trips in 15 American cities, coupled with a large TNC's survey of 4,365 of its users in late 2018. The TNC survey included stated preference questions focused on various alternative options for each participant's most recent trip choice. The TNC survey explored different market segments: for users who took a private TNC ride, would they be willing to choose a less expensive and longer ride, and for users who took a shared TNC ride, what changes would lead them to choose a more expensive and faster ride?

FHWA conducted a separate analysis surrounding users' preferences when choosing one shared TNC option over other shared TNC options. Results yielded significant heterogeneity in cost and time savings tradeoffs among users, meaning that there were substantial differences in user preferences. This suggests that, by offering customers more than one shared product option with time delay and varying price points, and by providing these in combination with vehicle routing decisions designed to accommodate differing user preferences, TNCs could increase the proportion of shared trips.

Ridehailing is not the only way to share private automobile rides. Some transportation agencies are working with emerging app-based carpooling and navigation services, like Scoop™, Waze™, Metropia™, Agile Mile™, and Hytch™, to pilot incentives for ridesharing in private vehicles. This study uses data from and analyses conducted by two app-based systems (Metropia™ and Hytch™) to understand the impact of incentives on desired behaviors.¹ These should only be considered as examples, since app-based carpooling systems that exist in the marketplace vary significantly in their user interfaces and the incentive structures they offer. Moreover, for both Metropia™ and Hytch™, the study used data that the app developer collected previously and not specifically for the research questions of interest to this effort. An important additional limitation was that travel choice data gathered via surveys prior to app use was not sufficiently specific to enable attribution of the cause(s) of behaviors to the use of these apps.

Based on the data and analysis gathered from the two sharing approaches and literature, the FHWA study developed an analytical model to facilitate the assessment of three scenarios (plus one experimental scenario) to test incentive impacts and the effectiveness of strategies to encourage travel choices that increase vehicle occupancy through sharing TNC rides and carpooling in private vehicles. Two scenarios focused on TNCs by increasing the cost savings of shared trips relative to private trips (cheaper) or by reducing the travel time penalty for shared trips relative to private trips (faster), one that focused on personal vehicles by increasing the price difference between private car trips and carpools, and an experimental scenario of rewarding shared personal vehicle trips through an app. Due to insufficient data regarding the impacts of incentives of interest, the last scenario was deemed experimental in nature. Currently, available data in this experimental area does not lend itself to a robust scenario evaluation. This study explains why such a scenario would be valuable and outlines the future research efforts needed to discern the impact of app-based incentives on carpooling formation.

Analysis of the scenarios reveal that changes to the relative price of private vehicle travel (i.e., driving alone rather than with a passenger) offer the greatest opportunity for reduction in VMT. The reason for this impact is that private vehicle travel accounts for the majority of VMT and person trips in the United States. The study finds, for example, that a $1 per trip increase for "drive-alone" trips over the price for shared trips would save more than 3.5 billion vehicle miles annually in the 15 markets studied in this report. More modest or more targeted interventions could also reduce VMT by focusing on particular geographies or population segments.

This study does not explore factors (beyond cost and travel time) that, according to other research, sometimes make people averse to sharing a vehicle with strangers, such as safety, privacy, and convenience. Secondly, analyses in this study do not address interactions across all modes. For that reason, this study can estimate how price and time affect a user's choice between a private and shared TNC ride, but it does not estimate how price and time affect a user's choice among TNCs, transit, driving, carpooling, walking, bicycling, or any other modes. Similarly, the study provides no data to consider how TNC characteristics affect a user's decision to take a trip in the first place. Additional multimodal discrete choice analysis is necessary to properly nest these decisions within an integrated mode choice model.

Table 1 organizes findings from this study by research question and directs readers to sections of this paper with related key findings.

Table 1. Summary of findings.

Research Question	Key Findings	Technical Details
For which types of trips are transportation network company (TNC) users most willing to share rides?	Solo trips, weekend trips, trips from work, trips home, and trips to entertainment or personal business were a greater proportion of shared trips.	Chapter 2: Distribution of Shared and Private Trips by Trip Characteristic
How much cheaper would the price of shared TNC rides have to be than private TNC rides to increase the probability for ridesharing by 10 percentage points?	A price difference of $1.16 per mile would increase the probability of sharing for general trips by 10 percentage points (from roughly 30 percent of trips to roughly 40 percent).	Chapter 2: Effect of Price on Sharing
How much less travel time penalty for shared TNC rides relative to private TNC rides would be required to increase the probability for ridesharing by 10 percentage points?	A travel time difference of 18 seconds per mile would also increase the probability of sharing for general trips by 10 percentage points (again, from roughly 30 percent of trips to 40 percent). This differential might be difficult to obtain from travel time changes and could more easily result from changes in waiting time (e.g., by prioritizing pick-ups for shared TNC rides).	Chapter 2: Effect of Time on Sharing
What types of TNC trips are relatively insensitive to differences in price between shared and private rides?	Trips to or from the airport or other intermodal travel centers were very insensitive to price, likely due to time sensitivity. Similarly, trips paid for by a third party were also insensitive to price differences.	Chapter 2: Market Segmentations and Price Sensitivity
What proportion of TNC users are completely unwilling to consider taking a shared ride?	Almost 35 percent of TNC users would not share a ride for any price savings presented, even without any time penalty.	Chapter 2: Effect of Price on Sharing
For which sorts of trips are TNC users most sensitive to differences in travel time between shared and private rides?	Trips to, from, or within dense office districts or areas with competitive transit, or trips to or from the airport or other intermodal travel centers were sensitive to differences in time, and thus were less likely to be shared.	Chapter 2: Relative Effect of Price and Time
What personal characteristics make a TNC user more likely to select ridesharing, based on other research?	Frequent sharers were more likely to be younger, unmarried, female, from zero-car households, and more frequent transit users.	Chapter 2: Ridehailing Versus Ridesharing
Could heterogeneity in user time and cost tradeoffs among shared product offerings enable TNCs to design additional such offerings that could lead to more sharing?	The high degree of heterogeneity in user preferences found in the study suggests that offering customers more than one shared product option with time delay and price points that vary could, if done in combination with vehicle routing decisions, increase the proportion of trips that are shared.	Chapter 2: Relative Effect of Price and Time
What factors limit carpooling adoption?	Convenience and trust issues are the two most critical factors that limit carpooling adoption.	Chapter 3: Overall Context
How effective are carpooling incentives in sustaining desired behavior among users of transportation apps that provide such incentives?	Incentives of only 2 cents per mile on the Hytch™ platform were effective at sustaining desired travel behavior. Monthly average awards of $7.54 for participants receiving 2 cents per mile appear affordable when compared to other transportation investment options. Both Metropia™ and Hytch™ gradually reduced reward levels over time, with user engagement tending to hold stable or grow despite the reductions (although it plummeted when the elimination of rewards was tested).	Chapter 3: Impact of Rewards and Incentives using Examples from Hytch™ Chapter 3: Impact of Rewards and Incentives using Data from Metropia Driving Up Occupancy Platform