Tensions and complementarities in mass transit and ride-hailing decisions through a survey-based randomization

In this paper, we focus on the effect of changes in the per-mile price of ride-hailing on various types of transportation modes. Here we use “price” to indicate the per-mile price charged, whereas we use “cost” to indicate the resulting cost of the full trip. Because the low per-trip cost has been a major driver in the success of TNC’s and ride-hailing is anticipated to become cheaper in the future, we take a different approach from previous studies by exploring possible future scenarios in which the per-mile price of ride-hailing is lower relative to today. In order to determine the causal effect of a per-mile price decrease, we implement a survey instrument designed as a randomized controlled trial. We specifically examine the effect that a decrease in the per-mile price of ride-hailing has on short-run, daily mode choices for primary commute trips. Each survey participant is randomly assigned one of three ride-hailing per-mile price points: $0.20 per mile, $0.70 per mile, and $1.2 per mile (all of which are typically lower than the current estimated average charge of $1.50-2.10 per mile ).
Participants are told to think about a hypothetical world in which they have recently learned that the price per mile of ride-hailing service such as Uber or Lyft is now a given value. To make this more tangible, we also give an estimate, given that hypothetical per-mile price, of the total cost of a ride-hailing trip to get to their specified primary destination (we ask for their most frequent primary destination earlier in the survey) on a typical day. We then elicit their stated preferences for the type of mode trip they would take to their primary destination on a typical day given this per-mile price. Specifically, we ask them to indicate whether their trip would be single-mode (taking one transportation mode for the entire trip), or multi-mode (taking more than one transportation mode in combination), and to specify which transportation mode or modes are included in this trip

2018-01-01

2018-06-30

We specifically examine the extent to which ride-hailing substitutes for the use of personal vehicles, and whether it will substitute or enable various forms of transportation.

Experimental Procedure. Our survey instrument asks several questions about transportation behaviors, attitudes, and preferences; in this paper we focus on a question that asks about hypothetical transportation choices given a specified price per mile of a ride-hailing trip. This question is designed as a randomized controlled trial. We define the following transportation modes: (i) personal vehicle, (ii) ride-hailing, (iii) bus, (iv) other mass transit (e.g., train or ferry), (v) walk (more than 5 minutes) or bike, and (vi) other. The survey asks participants what their primary destination is on a typical day. We know the respondent’s home address, so using the Google API we can calculate the distance between their home and primary destination for their most common commute trip. We randomly assign each survey participant to one of the following hypothetical prices for ride-hailing: $0.20 per mile, $0.70 per mile, and $1.2 per mile. Given this per-mile price, we provide an estimate of the whole cost of a ride-hailing trip, taking ride-hailing all the way from their home to their primary destination. We ask which transportation mode or modes they would take on a typical day, and have them indicate through a check box whether they would take each mode for the whole trip or for part of the trip (see the Appendix for the relevant survey questions as posed to participants).
These three prices were chosen to provide strong variation in choices. The $1.20 per mile price point tends to be less than the current cost of a typical UberX trip, which is usually closer to $1.50-2.10 per mile depending on the trip length. We imposed variation in the range below current prices to understand the implications of sizable drops in per-mile prices that might result from large scale-ups in service, automation, or other significant technological or business innovations that would substantially reduce marginal costs of supplying trips. We designed the study to explore future scenarios when ride-hailing could become much more inexpensive in order to ascertain what impacts to the transportation system we would face in such scenarios.
The question has a logic check in it that either allows participants to choose one mode choice for their whole trip, or two or more mode choices for part of the trip (but does not allow the participant to check two boxes in the whole trip column; check one in the whole trip column and one in the part trip column; or check only one in the partial trip column).

Computer

individual

No

800 people

800 people

200 low price, 200 medium price, 200 high price, 200 medium price but uncertain

have sufficient power to identify at least a 11-23% effect between treatment groups with alpha=.05 and power level 0.8, depending on the hypothesis being tested.