Alon Levy and I are having a pleasant discussion about the pros and cons of American-style commuter rail systems.
One characteristic of most American commuter rail networks is that they are “peaky”; that is, there is a substantial difference in service frequency between peak and off-peak hours. Alon sees peakiness as prima facie evidence of underutilization, something to be avoided. So let’s explore some of the underlying issues that lead to “peaky” transit.
Low density leads to peaky transit
The Main Line to Paoli used to have 15-minute, rapid-transit type service up through the 1970s. The infrastructure is still there to support it, but off-peak service today is every 30-40 minutes. Two things happened there. First, Center City declined in regional importance even as it grew in absolute terms, as places like King of Prussia and Great Valley siphoned off commuters. Second, a large amount of low-density residential development occurred which was not near Philly’s historic rail lines. A 1950’s student commuting to Bryn Mawr/Villanova/Haverford/St Joe’s was likely to live near a streetcar or regional rail line, enabling an all-transit trip. A present-day student is not.
In matrix form, the transit-density relationship looks like this:
Causality runs in all directions. Just as a decline in (relative) density moved the Main Line away from rapid transit, so has the gentrification of Chicago’s North Side moved that area toward more frequent service. For a long time, the Ravenswood L terminated at Belmont on nights and weekends, but in 2000 it was extended into the Loop at all times. Platforms were lengthened from 2009, allowing 8-car trains. More service was added this year, and a future project will add a flying junction to send northbound Ravenswood trains over the top of the Evanston/Howard tracks.
Likewise, low density areas rarely warrant rapid transit style service. In many cities, this leads authorities to propose extensions using different technology, with a forced transfer. Examples include eBart, the Denton A-train, and diesel trains that end at Fannin South.
Seating riders leads to peaky transit
Un-peaky, rapid transit-style services typically see a wide fluctuation in car occupancy. A DC Metro train at late morning may have only 30% of the seats taken, while one operating at the peak will be crammed to crush capacity. The same holds true for the NYC Subway, BART, and the Tokyo suburban network.
Alon lists a 2:1 ratio of peak-hour services to midday services as his ideal upper bound for rail. Most rapid transit outline vehicles have a crush capacity of about 2.5 times their seated capacity. A system with a 2:1 peak service ratio which runs at crush during the peak is thus carrying a 5:1 passenger ratio.
But wait, as the late Billy Mays would say – there’s more. Most subway/metro type services aren’t optimizing midday trains for full loading, they’re meeting a minimum performance benchmark – typically every 12, 15, or 20 minutes. Midday service would not be reduced unless the trains were noticeably empty.
A more realistic system, then, is one that sees a peak-hour loading factor of 2.25 (the threshold at which the operator would consider platform lengthening, reliever routes, etc) and an off-peak loading factor of 0.30 (the threshold at which the operator would consider reducing service).
At a 2:1 service peak, this line is carrying a 15:1 ridership peak.
All well and good, for a metro. But let’s change a couple of our assumptions. Let’s first imagine a system where commutes are very long – say, over 60 minutes. Riders with such commutes immensely value the ability to remain seated, as mentioned by commenters here and here. If we design the service so that peak loading is 1.0 and off-peak is 0.3, our 15:1 ridership peak requires a 4.5:1 service peak.
Let’s also suppose this train consists of loco-hauled diesel trainsets, which are fuel hogs. On an unelectrified line, there is a substantial financial incentive not to run trains at 30% capacity. You might instead consider a service cutback at 75% peak occupancy. The required service ratio in this instance is 12:1, which is well into Metra territory.
This is why brand-new commuter rail systems are more likely to emulate Metra’s peaky service than Paris’s RER. SLC’s Ogden-to-Provo Frontrunner – which largely operates over its own track – drops to hourly at midday. New Mexico’s Rail Runner has no midday service on weekdays, despite zero freight interference. (BNSF actually wants to abandon the Raton Pass line now that they’ve double-tracked Abo Canyon).
Transit bloggers like to imagine that if you provide frequent all-day service to the ‘burbs, you will open up a substantial new market of inter-suburban trips. However this ignores the fact that most of those people will just drive. And while institutional inertia can explain service provision on Metra or the LIRR, it doesn’t explain why New Starts systems, unencumbered by freight or legacy bureaucracy, nonetheless trend towards the same operating patterns.
Peaky transit is self-reinforcing
One of the characteristics which subways and freeways share is that they tend to prod people toward alternate work schedules. The early birds get up even earlier to “beat” the traffic; the late-risers negotiate a delayed start and go home when the worst of the afternoon rush has passed.
Beyond the unpleasantness of a packed subway car or a jammed highway, undifferentiated services are also faster in the off-peak. Dwell times are shorter, average speeds are higher. By contrast, peaky services actually get faster during the peak, because the ridership spike leads to different service patterns.
Let’s take Metra’s UP-West line as an example. Most of the morning service follows a “two-step” pattern in which trains originating from the outer ‘burbs run express from Elmhurst or Glen Ellyn and local trains slot in behind them. But with the 8:05am departure from Elburn, trains make all stops. What was formerly a pleasant 41-minute ride from Wheaton now takes a full 56.
UP-North has an even smaller express window. The first proper express out of Waukegan doesn’t arrive at Ogilvie until 8:19am; the last arrives at 9:15. Thus does the Metra timetable illuminate suburban class differences. The proles out in Kane County are piling onboard at zero dark thirty for their eighty-five thousand, while their boss in Lake Forest is arriving at 8:45 like a civilized human.
For my money, though, the peakiest express in the US has to be on SEPTA’s Paoli/Thorndale line, the artist formerly known as the R5. The Great Valley Flyer, trains 9526/9561, runs once a day, pulling into Suburban at 7:51 and departing at 5:08. If you can swing it, this train cuts a full 26 minutes off the all-stops local. I haven’t ridden it, but I know a guy who regularly makes the Exton/Center City run, and he says it’s the cat’s pajamas.
Whom shall we serve?
Since rapid transit-level service can only exist in high-density corridors, we’re left with a few options.
(i) Allow the city to grow as it will, and only serve the high-density corridors. Houston’s light rail has taken this approach, although it is mitigated by the existence of an extremely peaky commuter bus network. This helps promote a core of hip, walkable neighborhoods, but leads to a lot of auto-based reverse commuting.
(ii) Attempt to restrict low-density development, nudging everyone towards high-density living. Portland and Vancouver have followed this approach. While this “works” from a ridership perspective, it enrages the red tribe, leading to incoherent George Will columns and pushing growth into more accommodating jurisdictions.
(iii) Attempt to serve everybody with an overlapping set of peaky and non-peaky services. Salt Lake (Trax/Frontunner), Seattle (Link/Sounder), and Los Angeles (Metro/Metrolink) hew to this format. Vancouver gestures in this direction with the West Coast Express, although they really ought to have some sort of Abbotsford/Langley service.
(iv) Build a fast LRT/S-Bahn service that hits 80% of the metro but leaves out the outer-outer ring. Dallas is there; the trains get you to Plano and Rowlett, but there’s nothing in Frisco or McKinney. Denver will be soon; the LRT hits most southern suburbs, but permanently leaves out Castle Rock and Colorado Springs.
All of these options have pros and cons. I have a strong personal preference against (ii), since I believe people should be able to live how they like. Houston has made (i) work, but this is largely because we built out the bus network before we started on rail. If Houston had instead tried to build out an “express streetcar” LRT and outer-suburban HOV at the same time, there would’ve been endless debates of the “why are we spending money on buses to the suburbs when we could promote walkable development” sort.
(iii) and (iv), then, seem like the best options for most North American cities. That doesn’t mean there isn’t room for improvement on legacy systems. Additional Metra-CTA connectivity would be good, although I think the Chicago transit geeks are needlessly dismissive of the transfer opportunities that already exist. And the LIRR has some downright questionable operating patterns which Alon has devoted many pages toward improving.
But if we’re going to allow people to live in low-density suburbia, we ought to make peace with peaky transit. Whether that takes the form of the Great Valley Flyer or the West Bellfort conveyer belt matters not.