April 20, 2021

How does ridehailing use affect vehicle miles traveled?

Sustainable Transportation Lab alum and UC Davis PhD Student Summer Wu and I have published another new paper on the effects of ridehailing services in the US. Like our other recent paper, this one relies on data from the U.S. National Household Travel Survey, but now we focus on how the use of ridehailing services affects the overall amount of vehicle miles traveled (VMT) in the US. VMT is a measure of the overall amount of vehicle travel. All else equal, more VMT translates into more traffic and more emissions.

Our top line result, amplified on Twitter by CityLab writer David Zipper, is that at the time of the 2017 NHTS, ridehailing services were likely generating an additional 7.8 million vehicle-miles each day across the U.S. That sounds like a big scary number, but it represents an increase of about 0.1%. And it misses some fascinating and important nuance.

In my view, the far more interesting results had to do with comparing non-users, occasional users, and frequent users of ridehailing services. The problem with comparing these groups, of course, is that they are different in other ways besides ridehailing usage – age, income, employment status, where they live, and more – and those differences themselves lead to different levels of per-person VMT generation. Therefore, we created matched sets of non-users, occasional users, and frequent users, that were similar in their socio-economic characteristics and lived in similar neighborhoods within the same metro area. This makes for more of an apples to apples comparison. Furthermore, we analyzed separately (1) those who were licensed drivers and had access to a car, and (2) those who lacked either a license or a car. The figure below summarizes our key finding:

In each plot, the grey line represents a simple comparison between levels of use. That is to say, the values on the grey line are simply the values of VMT attributable to the average frequent, occasional, and non-user of ridehailing. The red and blue lines represent comparisons of like-with-like; each frequent ridehailing user in the sample has been matched with a demographically comparable occasional user and a demographically comparable non-user from the same Core Based Statistical Area. (The three different lines represent different assumptions about how many empty miles each ridehailing car travels for every mile of passenger-carrying service, ranging from 0.4 to 0.8 non-revenue miles per revenue mile.)

Interestingly, we can see that among licensed car owners (the left graph), occasional ridehailing users generate more total VMT than comparable non-users, but frequent ridehailing users generate the least of all. We suspected that this could be a sign of occasional users using ridehailing as an additional mode in addition to car ownership, and frequent users turning ridehailing into more of a replacement for car ownership. And indeed, we found that frequent ridehailing users owned an average of 0.2 fewer cars than demographically comparable non-users.

Unsurprisingly, for those lacking a driver’s license and/or a car (the graph on the right), more frequent ridehailing use translated directly into more VMT. However, the overall amount of VMT generation was much lower in this group than for the licensed car owners, regardless of the level of ridehailing use.

Proponents and detractors of ridehailing services continue to debate over whether ridehailing reduces automobile travel by providing an alternative to car ownership, or exacerbates and perpetuates the problems caused by car dependence. In short, we find evidence supporting both of these narratives, for different groups of people. We conclude our paper with this observation:

One potentially important implication of this is that if policymakers aim to reduce VMT by turning car drivers into multimodal travelers, they should focus on trying to encourage frequent rather than occasional RS use by car owners. Converting car owners to occasional RS users appears to increase VMT, but taking them all the way to frequent user status leads to lower VMT generation. This may be accomplished by targeted incentive/disincentive transportation demand management programs. At the same time, as our finding suggests, policymakers may not want to encourage non-car-owners to become frequent RS users, as RS services uniformly increase VMT for this group.