Lyft Shuttle doesn’t solve congestion, it creates it

Recently, Lyft launched what they call “Shuttle” and what everyone else calls “the bus.” It is an exceedingly inefficient bus, given that it has a maximum occupancy of 4 (but an average of likely between 1 and 2). Everyone made fun of them, so they wrote about how, no, it’s actually great, because it’s not a bus, even though it acts like one, and it will reduce car ownership and expand mobility.

And what they wrote is, well, kind of infuriating, but a good lesson as to how ride-hailing, even if it’s one day autonomous, will not solve congestion, and may just make it worse. It shows that Lyft not only seems to have no understanding of how transit economics work, but also no understanding of how their own service works, assuming that vehicles can just appear at random to fulfill demand. Their argument begs to be broken down line by line, so I’ll do that. Original indented, my comments in italics.

Recently, Lyft started testing a new product called Shuttle in two cities: Chicago and San Francisco. I’ve seen quite a few thoughtful questions — and a few good jokes — circulating about Shuttle. So as one of Lyft’s longtime resident transit nerds, I want to take the time to share why we launched Shuttle, and what it means for cities and the future of public transportation.

Okay, so you’re a transit nerd. Great! Let’s see how well you understand transit.

What is Shuttle? 

You can think of Shuttle as a new twist on Lyft Line, the popular dynamic carpooling option we launched in 2014, which is available in 16 cities and accounts for around 40% of all rides in those markets. Just like Line, Shuttle matches you up with other passengers going the same direction to create a high-occupancy ride for a lower price than riding alone (more people in fewer cars = win). The twist is that unlike Line, Shuttle pick-up and drop-off locations are the same every day, and so are the prices (with prices between $3–4 per ride, Shuttle is Lyft’s most affordable service, offering commuters a reliable option for far less than the average cost of car ownership, which is about $9,000 per year). And instead of coming straight to your door, Shuttle picks you up at a safe, carefully selected spot nearby.

The main quibble here is the $9000 cost of car ownership. That assumes that you live in the suburbs and actually drive your car all the time! But the rides described here are generally 2 to 5 miles. Let’s assume 5 miles, both ways, 250 work days per year. That’s 2500 miles. Yet the $9000 figure is based on 15,000 miles per year. If you don’t have a car, Lyft will save you all the fixed costs of ownership, but if you want to go somewhere outside of commute times, you have to pay the full cost per mile in rental fees and short-term insurance. (Or take a Lyft/Uber, but that’s relatively expensive, too.) And if you do have a car, Lyft is only saving you the marginal costs of about 15¢ per mile, which for 2500 miles comes out to just $375, but you’ll pay $2000 for the Lyft rides (80¢ per mile, which is quite a bit more, per mile, than car ownership). You won’t, at least, have to pay for parking.

As many on social media pointed out, these new features are also found in traditional public transit buses. Fair enough. But here’s what’s different and important. 

Shuttle rides are provided by the same, awesome Lyft drivers you know and love, driving their own cars — not big vans or buses.

This is a bug, not a feature! The reason why transit works economically is because it uses larger vehicles: it can scale by carrying more passengers with the same number of drivers. Sure, a 40-foot transit bus isn’t efficient if it’s carrying two passengers, but in a city like Chicago or San Francisco, it’s not. In the corridors Lyft Shuttle serves, buses in the peak are generally full. That’s much more efficient than a car, even one carrying three passengers.

As a driver in a Shuttle market, your next ride request could be an original Lyft, Line or Shuttle request; they’re all part of the mix. That means no one needs to go buy any new vehicles to turn on a Shuttle route, and they are easy to change based on feedback. And unlike a bus route, a Shuttle route is on-demand — it doesn’t happen unless a passenger requests it.

So are bus routes. None of the routes in the Shuttle areas have empty buses (or trains). In fact, the buses there are well-used enough that frequencies are based not on some minimum level of service (a bus every 15 minutes, for example) but to meet passenger demand, some as frequently as every four or five minutes. 

That means Shuttle is very efficient, never making drivers circle and waste miles when people aren’t riding it. Instead, drivers just move on to give other Lyft requests after they finish a route, or log out if requests slow down.

This completely misses the fact that cars don’t appear and disappear but have to come from somewhere and go somewhere before and after a trip, usually not carrying passengers. If a driver gets downtown and drops off a fare and finds there isn’t much demand (since at morning peak hour, there is far more demand going to downtown than leaving, with the opposite in the afternoon) the car doesn’t just go *poof* and disappear, but it has to be driven somewhere else. The problem here is two-fold: first, this creates additional congestion, since transporting passengers for 4 miles might require the car to drive an additional 4 miles in an already congested area without any passengers. If the system were perfectly balanced with trips into and out of the core, it might work, but in San Francisco and Chicago, at rush hour, this is certainly not the case.


Second, the driver of this vehicle is not compensated for this time, so it eats in to their already-low wage. This speaks to an obvious flaw of the system: at the current price of $4 per ride, it’s completely unsustainable. Let’s assume that Lyft carries an average of 3 passengers per trip at $4 fare each. That’s $12. Lyft takes 20% off the top, so the driver is getting $9.60. The marginal cost of a 35-minute, 4 mile trip in rush hour is probably $1 for gas and maintenance, but then the driver has to deadhead (operate without a fare, while still incurring the costs of operation) to and from the next fare, which likely adds another 4 miles ($1) and 15 minutes. Now you’re earning $7.60 for 50 minutes of work, or $9.12 per hour. And that’s sort of a best-case scenario, and assumes you have relatively little time in between fares, and doesn’t account for time spent pumping gas, cleaning the vehicle, and other items. If we assume that every other trip, a driver will have no fare and have to deadhead 8 miles ($2, 30 minutes) out of the city, the economics collapse: the average hourly wage falls to $6. And while there is some affordable housing within 8 miles of Chicago, there certainly isn’t that close to San Francisco.  

This form of flex transit capacity augments the fixed capacity of traditional transit routes at peak times when those services are maxed out with riders, brings “right-sized” service into areas which can’t support traditional transit, and melts away when it isn’t needed.

This sentence is trying to have it both ways! If traditional transit routes have fixed capacity, it means that they are full, and if they are full, it means that it doesn’t serve an area that can’t support traditional transit. None of the corridors described are in areas that don’t support traditional transit, of course. And then the “melting away” phrase doesn’t account for the fact that it would wind up with a bunch of empty vehicles driving away from downtown, adding to congestion and pollution.


This is the problem with ride-hailing and (potentially eventually) autonomous vehicles. It may reduce the number of parking spaces we need, which would be a good thing. But unless the system is perfectly balanced—and in cities, it inherently is not—parking will be traded for more vehicle miles traveled (VMT). More VMT means more congestion. Is this a good trade? I’m not so sure. 


This is a problem in the transit industry as well: full buses come downtown, drop off passengers, and operate empty to a garage to sit idle for the middle of the day. But it’s an order-of-magnitude bigger problem for smaller vehicles: a bus can carry 60 passengers and make one empty trip. Cars—even with three passengers each—would require 20 such trips with 20 outbound drivers.

Shuttle’s initial routes were chosen by our data scientists, who looked at the density of Lyft Line requests during commute hours and picked a small set of initial corridors where the data showed we would be able to match up the most passengers, enabling the highest occupancies and thus the lower prices passengers want.

Or the low prices which can entice people into these cars. As mentioned above, these prices are nowhere near sustainable. Plus, passengers don’t really want high occupancy. If they did, they’d ride the bus; if they didn’t, they’d drive instead. Four passengers means someone has to sit in the middle seat, which is likely less comfortable than a bus line. (And on a bus, you can at least move to another seat or part of the bus if you’re uncomfortable with the person next to you. Try that in the middle seat of a Corolla.) If Lyft regularly has four passengers per driver, customers will abandon the service. The practical maximum number of passengers is 3, making the average less, making the math shown above a rosy estimation.

Not surprisingly, many of these corridors align with those most heavily used by other transportation modes, including personal cars and transit. With Shuttle, we hope to reduce the need for commuters to drive their own cars on those busy corridors, giving them a more efficient option and freeing up road capacity for high-efficiency shared and transit vehicles.

This is a nice sentiment, but it only works if significantly more people switch from driving to Shuttle than from transit to shuttle. Let’s imagine a congested corridor with a bus carrying 100 people and 100 cars carrying one person. In many cases, the corridors Lyft Shuttle is serving have 2-to-1 transit-drive mode share (and that’s for all trips, it’s probably even higher for trips to the core, see this map of Chicago, for instance). If 45 riders take shuttle at a 2-to-1 transit-to-driver ratio, this would add 18 vehicles to the roadway while removing 15 cars, meaning it would be less efficient and would certainly not free up road capacity. (This assumes zero deadhead mileage, which adds further congestion while providing no benefit.) Given recent trends of ridership declines on transit and increased congestion, Occam’s Razor would suggest that’s exactly what happens.

We’re For Transit  

You might have noticed that Shuttle sounds like some hybrid mashup of ridesharing and transit. You’re not wrong. It fits into a new category experts call “microtransit,” and it’s designed to do things buses can’t do and reach people buses don’t reach, helping attract a broad spectrum of riders who haven’t used transit before.

Microtransit doesn’t really work. Bridj failed to make money or expand beyond Boston, even with larger vehicles than Lyft Shuttle and fares in the $5-$6 range, except for a publicly-funded foray into Kansas City. In that case, the year-long pilot project had 1480 riders. Not per day, total. At a cost of $1.3 million, it was a subsidy of $878 per ride. That pilot attempted to do things buses don’t, reach people buses can’t and bring in new transit riders, and it didn’t exactly work. The most successful microstransit provider, Chariot, runs a few fixed route transit-style services in San Francisco, but mostly serves corporate-subsidized routes to suburban campuses. 

The Lyft team is made up of strong public transit advocates (some might say geeks), a tradition that started with our co-founder and CEO, Logan Green.

Remember when your middle school teacher—and Hemingway, apparently—would say “show, don’t tell”? It’s a rhetorical tell when you have to open every paragraph by telling us how you really, truly love transit. 

We come to this issue with a deep empathy for and exposure to the challenges facing traditional public transit operations. Over a decade ago, Logan served on a public transit agency board in Santa Barbara, California. He lobbied hard to improve the frequency of bus service in the city’s low-density neighborhoods to benefit riders who depended on buses, only to find that the transit agency couldn’t afford it. There just weren’t enough riders to fill more buses on those routes.

Yet Lyft Shuttle serves areas so dense that the only reason they can exist is because of transit. If Santa Barbara needs Lyft Shuttle so much, maybe it should try to operate there.

Never one to take no for an answer, Logan was inspired by the informal jitney networks he saw on a trip to Zimbabwe and became determined to find another way to expand mobility access.

Yes, since Zimbabwe is a perfect corollary for the Loop in Chicago. And in many cases, third-world jitney networks have been replaced with, you guessed it, transit.

His vision was to turn every single-occupant vehicle on the road into an extension of the transit network, which would cut traffic and make it easier for everyone to get around at a low cost — especially people who couldn’t afford to own cars. 

But Lyft and Uber don’t do that: they add vehicles to the roads and generally increase traffic in already-dense areas.

Shuttle is the latest step in making that vision happen. In parallel, we’re also continuing our strong policy advocacy for expanded investment in traditional public transit, such as Measure M in Los Angeles, Proposition 1 in Seattle, and Measure RR in the Bay Area, all of which were passed by voters last fall. Our country must break down the traditional binary between cars and transit and embrace all of these strategies if we want to make gains in reducing reliance on car ownership.

The whole point of not owning a car is to not drive a car.  Or if you don’t you want to save the money you spend owning a car and dealing with the hassle? Ride in a car but have a chauffeur, we’ll explain how for some reason that’s cheaper. (It’s not, except in small doses.)

How Shuttle fits in with Transit  

New, creative and efficient ways of delivering transit are sorely needed. For over a century, public transit agencies have had two basic tools for getting people around cities: buses and trains. Like any tool, buses and trains work really well for some tasks, and not very well for others. For moving large volumes of people along heavily-traveled corridors during commute times — buses and trains can’t be beat.

So, you mean, basically every corridor Lyft Shuttle operates in. Downtown San Francisco has more transit commuters than drivers from the city. The corridors in Chicago are similar: for travelers to the Loop, 189,000 take transit, 81,000 drive.

They move the most people at the lowest cost while taking up the least space on the road — IF they are able to attract enough riders.

Again, this doesn’t seem to be a problem in Chicago or San Francisco.

But for routes where there are smaller numbers of riders, traditional buses and trains struggle to provide frequent, cost-efficient service, and ridership suffers accordingly, leaving riders in those areas with few efficient options. We think Shuttle can help bridge these gaps, and it’s one reason why we launched an initial test route in a transit-underserved area on Chicago’s South Side that connects to Downtown from Brighton Park. 

That pilot would be useful if there weren’t any transit options from Brighton Park to the Loop. But there are: the Orange Line or the six bus lines which serve the neighborhood every 10 minutes or better at rush hour.

The vast majority of Americans live in places outside of mass transit’s “sweet spot” — which is why only about ⅓ of jobs in major U.S. metropolitan areas are reachable by the average commuter within 90 minutes one-way by public transit.

But downtown San Francisco and Chicago are not exactly transit deserts.

And many potential transit riders live too far away from train stations to be able to use them, a classic challenge known by transit agencies as the first-and-last-mile problem.

What about bus stops? Buses are bad, but using smaller cars as buses is great?

Here again, we think Shuttle can help. Since launching Shuttle in San Francisco, we’ve found that our single most successful route is a first-and-last-mile connector to the 4th and King Caltrain station.

Then run a last-mile shuttle. Or advocate for bus lanes on streets from 4th and King to Market Street (but that would preclude Lyft pulling over in said transit lanes). But if Lyft Shuttle is going to reduce traffic, it’s not going to do so by acting as a last-mile one-to-three occupant shuttle service from 4th and King. Thousands of people arrive by Caltrain at peak hour in San Francisco, there’s just not enough road space for them all to get into cars for the last mile. How many of these Lyft Shuttle riders are switching from a single occupancy vehicle? Probably very few. Lyft Shuttle would probably work well for reverse-commuters way out in the suburbs, but the economics there are quite a bit harder, and it may be better solved by casual car-sharing. 

Public transit services also experience challenges on the flip side of the ridership equation. On the busiest transit routes, demand often exceeds supply at peak times, resulting in transit facilities which are severely over-capacity, unable to serve all the riders who want to use them.

But you said, three paragraphs ago, “For moving large volumes of people along heavily-traveled corridors during commute times — buses and trains can’t be beat.” So the solution is obviously to, uh, add more congestion to these corridors?

And when unusual events like severe weather, traffic crashes, or service outages occur it is hard for transit systems to quickly adapt to add more service in time to meet the need.

If transit is completely full, then, yes, it is hard to add more service. But if you need a lot more service, enough cars to provide that service will just exacerbate traffic.

This problem has been exacerbated by chronic underinvestment in funding for transit infrastructure.

Ding ding ding! That’s a real issue. Creating congestion and making transit less efficient isn’t going to help.

Flexible microtransit services like Lyft Shuttle and Lyft Line bring additional capacity into the system when and where it’s needed most, closing the last mile gap, delivering right-sized mobility service in underserved communities, and supplementing crowded public facilities. 

First of all, if Lyft is trying to bring additional capacity into the system, it is causing congestion, which slows down transit, and reduces capacity. Low-occupancy vehicles—which a Lyft Shuttle is, compared to a bus, even if it has two or three riders—do not add capacity to an at-capacity transit system: more buses do, or the same number of buses moving faster. And Lyft Shuttle can only do so many things, but bringing mobility to underserved communities and supplementing crowded public facilities are mutually exclusive. Right now, Lyft Shuttle does the latter, to the detriment of said facilities.

The bottom line is that in a country where fewer than 5% of commuters currently use transit, where 76% of Americans drive alone to work,

Except in areas served by Lyft Shuttle, the numbers are more like 25% car, 55% transit, with the balance walking and biking. When you can figure out how Lyft Shuttle can serve exurb-to-exurb commuters better than cars, you might have a point. But that’s a little harder.

and where traffic congestion continues to drag down economic productivity and quality of life,

This sounds like the old joke of a doctor, engineer and lawyer arguing about whose profession was the oldest. The doctor said God creating Eve from Adam’s rib made his the oldest, the engineer counters that God created the world from chaos, making GOd an engineer. But the lawyer counters: “yes, but who created all the chaos?”


Who do you think created all the congestion? Hint: it’s not buses. It’s cars. Even if they’re Lyft Shuttles with two passengers. 

our cities need a new spectrum of efficient, affordable transportation options to make car ownership a thing of the past.

If no one owns cars, but everyone rides around in cars, the outcome is the same. The point of reducing car ownership is to reduce the number of cars on the road. Having cars clog up bus routes and bike lanes doesn’t exactly do that.

Shuttle is a new tool to help make that happen. We welcome your feedback along the way as we continue working to improve people’s lives with the world’s best transportation. 

[exhale]

Lyft is trying to have it both ways with the Shuttle service. Their thesis is that it will expand mobility and reduce car ownership, yet they are serving already transit-rich areas with low car ownership rates, and it only operates during peak commuting times when most people are riding transit already. They claim it’s efficient, but it’s only more efficient than driving alone (or riding alone in a Lyft, which—with the requisite deadhead movements—is actually less efficient and might be the outcome in a “shuttle” service anyway). But it’s still far less efficient than transit, a fact which they mention several times. If Lyft can target a corridor with low transit ridership and move people to Lyft Shuttles (carpools), it will be effective in reducing congestion. (This happens when you provide a good carpool lane and minimal parallel transit.) But in congested, transit-served areas, it will provide no added mobility and increase congestion.

That’s not the kind of thing we’d expect from a company that is made up of “strong public transit advocates” now, is it?

Witches and Networks

A couple items of note from the Boston Globe today (one pay-walled, one not):

1. This summer, we posted on the well-used but ill-marketed weekend train service to beaches on the MBTA’s Newburyport-Rockport line. This past weekend, and this coming weekend, the T is adding several trains on that line. Why? The line goes through (quite literally, in a tunnel) Salem, that of witch trials and general spookiness. Hallowe’en is big business in Salem, with the monthlong “haunted happenings” peaking on the last two weekends of the month. With roads closed and parking limited, the city actively promotes the use of public transportation, and the T obliges. (Oh, and parking gets a lot more expensive. Supply et demand.) With the city backing, the train schedule is more than doubled on weekend afternoons.

Obviously, there’s capacity on the line (which handles nearly no freight—the main choke points are Beverly Junction and the aforementioned one-track tunnel through Salem). Why is there no extra service in the summer? There are probably a couple of reasons. Salem wants as many people as to come as possible—at least before Hallowe’en evening when things can get rowdy. More feet on the ground means more dollars in the coffers. The beach towns are not likely pressing the T for more people to crowd the beaches. They’d be glad to charge drivers $25 to park, but only until the lots are full. Otherwise, there’s not a lot of money to be made on beach admission. Second, the witch tourism is planned far in advance and not very weather dependent (although it likely drops off in case of rain or—this year, apparently—snow). A rainy day does not see many beach-goers. Still, there’s no reason the T shouldn’t at least promote the beach service they have in the summer, even if it’s not to the extent of doubling spooky service.

2. There was a very interesting event on the T yesterday: a startup meetup in the last car of a midday Red Line train. (The Globe article is paywalled and I can’t find a free link. Update: here’s a free link) Apparently, a bunch of local startup firms all met on the Red Line, had short speeches between stops, and then networked. Great idea! As one of the top venture sites in the country (although lagging behind Silicon Valley) many of these types are already on the Red Line every day, so what a good way to get people together. Now, if only we could have rational development policies which did not underprice suburban office parks and pull these companies out to Waltham and Burlington, where innovators spend way too much time behind the wheels of their cars.

Which reminds me … this year, I’ve been tracking my travel. Every day, every mile, every mode. Seriously. I have a spreadsheet, and every month I post a “transport report” for that month. (Here’s January; you can click to subsequent months.) At first, I had no idea what I’d use the data for (other than fun charts. hooray fun charts!). But now I have some ideas:

v 0.1: Manual data entered in to a Google Doc, manual charts made in Excel
v 0.2: Once I figure out how to make PHP and JPGraph work (any help would be appreciated, seriously), have the charts update live from a Google Doc / database when the data is entered
v 0.3: Open this to other transportation geeks who want to provide this type of data, and see if it works—and what happens. When you think about every mile you travel and every mode, you tend to drive less.
v 1.0: Track transport and mode using smart phones (doable, I think, based on discussions from Transportation Camp last spring) and gather data, and develop a social network based on where people are and when. What do you do with these data? All sorts of things. Like:

  • Real-time updates about traffic and transit problems sent to people based on their past behavior (i.e., we know you take the Red Line or drive Route 2 between 8:15 and 8:25, but there’s a disabled train / traffic accident, so you should consider the #1 bus / going to Alewife instead)
  • Helpful hints to help people drive less. A system could weed out short trips which could be made by foot or bicycle. Or help people link trips together. Or even take shorter routes.
  • Let people meet others with similar commuting patterns. If enough people used the system, it might be able to pair drivers in to car pools (“we’ve determined that you and another driver have each driven alone from the same neighborhood to the same office park with 15 minutes of each other 23 of the last 27 days—here is a link to send them an email/fb message/tweet to try to form a carpool.”) or meet people on the train or bus. To make the solitary confinement of a crowded train less—well—solitary. This would be optional, of course, but could be very powerful
  • Provide really interesting data to planners on who is going where, when and by what modes.
Obviously, some work needs to be done. But it’s an idea. If you have any ideas for it, or want to be notified when I get the system off the ground (or even close to that phase) watch this page or email me at ari dot ofsevit at gmail.