Read on Twitter
I'm pleased to share my publication, "The Congestion Costs of Uber and Lyft," forthcoming in the Journal of Urban Economics!
I use a natural experiment to identify the impact of ridesharing on traffic speeds, and then discuss policy implications. 1/n
https://www.sciencedirect.com/science/article/abs/pii/S0094119020300899?dgcid=rss_sd_all#
I use a natural experiment to identify the impact of ridesharing on traffic speeds, and then discuss policy implications. 1/n
https://www.sciencedirect.com/science/article/abs/pii/S0094119020300899?dgcid=rss_sd_all#
(2) On May 9th, 2016, Uber and Lyft abruptly left Austin, TX after voters rejected an uber-sponsored ballot initiative regarding driver background checks. Voila! A natural experiment.
(3) Using RD and DID designs, I estimate modest increases in traffic speeds (~2-4%) following the departure of ridesharing.
(4) Back of the envelope calculations using setting-specific measures of the value of travel time suggest that the slowdowns induced by TNCs cost Austinites $33 to $52 million annually.
(5) How big are these cost figures?
1. BoE calculations suggest they're ~ the size of the consumer surplus from TNC ops in Austin (Cohen et al. 2016)
2. Irnix's (rather opaque) reports imply that these costs are 1-2% of Austin travel time costs, and 4-6% of congestion costs
1. BoE calculations suggest they're ~ the size of the consumer surplus from TNC ops in Austin (Cohen et al. 2016)
2. Irnix's (rather opaque) reports imply that these costs are 1-2% of Austin travel time costs, and 4-6% of congestion costs
(6) So if Uber and Lyft impose this (modest) cost, what does this imply for policy?
(7) Banning/restricting ridesharing (e.g. NYC) on congestion grounds would be a good idea only if the slowdowns produced by TNCs swamped the benefits brought by U&L. In Austin, this doesn't seem to be the case.
(8) What about taxation?
Congestion pricing is hard to implement in practice, so it might make sense to impose TNC-specific taxes if these companies operate at places/times with high marginal external costs of congestion, especially if these trips are elastic.
Congestion pricing is hard to implement in practice, so it might make sense to impose TNC-specific taxes if these companies operate at places/times with high marginal external costs of congestion, especially if these trips are elastic.
(9) The thing is, TNC trips /arent/ that elastic (Cohen et al, again). And my paper suggests that TNC externalities don't necessarily occur at high-cost times or locations.
This means that TNC taxation shouldn't be expected to efficiently deliver congestion improvements.
This means that TNC taxation shouldn't be expected to efficiently deliver congestion improvements.
(10) That's /not/ to say that we shouldn't tax Uber and Lyft. If you're a city planner looking for a politically expedient & efficient (read: Ramsey-esque) way to raise revenue, look no further than a ridesharing tax.
Just don't expect it to make much of a dent in congestion.
Just don't expect it to make much of a dent in congestion.
(11) Lastly, worth noting that this is /one/ study in /one/ mid-sized city. The results may not extend to megacities. But Austin looks a lot like a lot of other mid-sized sprawl cities, so these findings are probably best applied in the Phoenixes and Denvers of the world.
