Working on an emergency timetable for regional rail has made it clear how an environment of austerity requires tradeoffs that reduce efficiency. I already talked about how the Swiss electronics before concrete slogan is not about not spending money but about spending a fixed amount of money intelligently; but now I have a concrete example for how optimizing organization runs into difficulties when there is no investment in either electronics or concrete. It’s still possible to create value out of such a system, but there will be seams, and fixing the seams requires some money.
Boston regional rail
The background to the Boston regional rail schedule is that corona destroyed ridership. In December of 2020, the counts showed ridership was down by about an order of magnitude over pre-crisis levels. American commuter rail is largely a vehicle for suburban white-collar commuters who work in city center 9 to 5; the busiest line in the Boston area, the Providence Line, ran 4 trains per hour at rush hour in the peak direction but had 2- and 2.5-hour service gaps in the reverse-peak and in midday and on weekends. Right now, the system is on a reduced emergency timetable, generally with 2-hour intervals, and the trains are empty.
But as Americans get vaccinated there are plans to restore some service. How much service is to run is up in the air, as is how it’s to be structured. Those plans may include flattening the peak and going to a clockface schedule, aiming to start moving the system away from traditional peak-focused timetables toward all-day service, albeit not at amazing frequency due to budget limits.
The plan I’ve been involved with is to figure out how to give most lines hourly service; a few low-ridership lines may be pruned, and the innermost lines, like Fairmount, get extra service, getting more frequency than they had before. The reasoning is that the frequency that counts as freedom is inversely proportional to trip length – shorter trips need more frequency and shorter headways, so even in an environment of austerity, the Fairmount Line should get a train every 15 or 20 minutes.
In an environment of austerity, every resource counts. We were discussing individual trains, trying to figure out what the best use for the 30th, the 35th, the 40th trainset to run in regular service is. In all cases, the point is to maximize the time a train spends moving and minimize the time it spends collecting dust at a terminal. However, this leads to conflict among the following competing constraints:
- At outer terminals like Worcester and Lowell, it is desirable that the train should have a timed transfer with the local buses.
- At the inner terminals, that is South and North Stations, it is desirable that all trains arrive and depart around the same time (“pulse“), to facilitate diagonal transfers, such as from Fitchburg to Salem or from Worcester to Brockton.
- Some lines have long single-track segments; the most frustrating is the Worcester Line, which is in theory double-track the entire way but in practice single-track through Newton, where only the nominally-westbound track has platforms.
- The lines should run hourly, so ideally the one-way trip time should be 50 minutes or possibly 80 minutes, with a 10-minute turnaround.
Unfortunately, it is not possible to satisfy all constraints at once. In an environment with some avenues for investment, it’s possible to double-track single-track bottlenecks, as the MBTA is already planning to do for Newton in the medium run. It’s also possible to speed up lines on the “run as fast as necessary” principle to ensure the trips between knots take an integer or half-integer multiple of the headway; in our higher-investment regional rail plan for Worcester, this is the case, and all transfers and overtakes are tight. However, in a no-investment environment, something has to give. The Worcester Line is 90 minutes end-to-end all-local, and the single-track section is between around 15 and 30 minutes out of South Station, which means it is not possible to conveniently pulse either at South Station with the other commuter lines or at Worcester with the buses. But thankfully, the length of the single-track segment between the crossovers is just barely enough to allow bidirectional local service every 30 minutes.
No-investment and low-investment plans are great for highlighting what the most pressing investment needs are. In general Boston needs electrification and high platforms everywhere, as do all other North American commuter lines; it is unfortunate that not a single system has both everywhere, as SEPTA is the only all-electric system and the LIRR (and sort of Metro-North) is the only all-high-platform system. However, more specifically, there are valuable targets for early investment, based on where the seams in the system are.
In the case of integrated timetabling, it’s really useful to be able to make strategic investments, including sometimes in concrete. They should always be based on a publicly-communicated target timetable, in which all the operational constraints are optimized and resolved for the maximum benefit of passengers. For example, in the TransitMatters Regional Rail plan, the timed transfers at the Boston end are dealt with by increasing frequency on the trunk lines to every 15 minutes, at which point the average untimed transfer is about as good as a timed hourly transfer in a 10-minute turnaround; this is based on expected ridership growth as higher frequency and the increase in speed from electrification and high platforms both reduce door-to-door trip times.
The upshot is that austerity is not good for efficiency. Cutting to grow is difficult, because there are always little seams that require money to fix, even at agencies where overall spending is too high rather than too low. Sometimes the timetables are such that a speedup really is needed: Switzerland’s maxim on speed is to run as fast as necessary, not as fast as trains ran 50 years ago with no further improvement. This in turn requires investment – investment that regularly happens when public transportation is run well enough to command public trust.
Outside a city core with very high frequency of transit, say 8 minutes or better, bus and train services must be timetabled to meet each other with short connections as far as possible. Normally, this is done through setting up nodes at major suburban centers where trains and buses can all interchange. For example, see this post from six months ago about the TransitMatters proposal for trains between Boston and Worcester: on the hour every half hour, trains in both directions serve Framingham, which is the center for a small suburban bus system, and the buses should likewise run every half hour and meet with the trains in both directions.
This is a dendritic system, in which there is a clear hierarchy not just of buses and trains, but also of bus stops and train stations. Under the above system, every part of the Framingham area is connected by bus to the Framingham train station, and Framingham is then connected to the rest of Eastern New England via Downtown Boston. This is the easiest way to set up timed rail-bus connections: each individual rail line is planned around takt and symmetry such that the most important nodes can have easy timed bus connections, and then the buses are planned around the distinguished nodes.
However, there’s another way of doing this: a bus can connect two distinct nodes, on two different lines. The map I drew for a New England high- and low-speed rail has an orbital railroad doing this, connecting Providence, Worcester, and Fitchburg. Providence, as the second largest city center in New England, supplies such rail connections, including also a line going east toward Fall River and New Bedford, not depicted on the map as it requires extensive new construction in Downtown Providence, East Providence, and points east. But more commonly, a connection between two smaller nodes than Providence would be by bus.
The orbital bus is not easy to plan. It has to have timed connections at both ends, which imposes operational constraints on two distinct regional rail lines. To constrain planning even further, the bus itself has to work with its own takt – if it runs every half hour, it had better take an integer multiple of 15 minutes minus a short turnaround time to connect the two nodes.
It is also not common for two suburban stations on two distinct lines to lie on the same arterial road, at the correct distance from each other. For example, South Attleboro and Valley Falls are at a decent distance, if on the short side, but the route between them is circuitous and it would be far easier to try to set up a reverse-direction timed transfer at Central Falls for an all-rail route. The ideal distance for a 15-minute route is around 5-6 km; bus speeds in suburbia are fairly high when the buses run in straight lines, and if the density is so high that 5-6 km is too long for 15 minutes, then there’s probably enough density for much higher frequency than every half hour.
The upshot is that connections between two nodes are valuable, especially for people in the middle who then get easy service to two different rail lines, but uncommon. Brockton supplies a few, going west to Stoughton and east to Whitman and Abington. But the route to Stoughton is at 8.5 km a bit too long for 15 minutes – perhaps turning it into a 30-minute route, either with slightly longer connections or with a detour to Westgate (which the buses already take today), would be the most efficient. The routes to Whitman and Abington are 7 km long, which is feasible at the low density in between, but then timetabling the trains to set up knots at both Brockton and Abington/Whitman is not easy; Brockton is an easy node, but then since the Plymouth and Middleborough Lines are branches of the same system, their schedules are intertwined, and if Abington and Whitman are served 15 minutes away from Brockton then schedule constraints elsewhere lengthen turnaround times and require one additional trainset than if they are not nodes and buses can’t have timed connections at both ends.
Planners then have to keep looking for such orbital bus opportunities. There aren’t many, and there are many near-misses, but when they exist, they’re useful at creating an everywhere-to-everywhere network. It is even valuable to plan the trains accordingly provided other constraints are not violated, such as the above issue of the turnaround times on the Old Colony Lines.
I want to follow up on what I wrote about speed zones a week ago. The starting point is that I have a version 0 map on Google Earth, which is far from the best CAD system out there, one that realizes the following timetable:
This is inclusive of schedule contingency, set at 7% on segments with heavy track sharing with regional rail, like New York-New Haven, and 4% on segment with little to no track haring, like New Haven-Providence. The purpose of this post is to go over some delicate future-proofing that this may entail, especially given that the cost of doing so is much lower than the agency officials and thinktank planners who make glossy proposals think it should.
What does this entail?
The infrastructure required for this line to be operational is obtrusive, but for the most part not particularly complex. I talked years ago about the I-95 route between New Haven and southern Rhode Island, the longest stretch of new track, 120 km long. It has some challenging river crossings, especially that of the Quinnipiac in New Haven, but a freeway bridge along the same alignment opened in 2015 at a cost of $500 million, and that’s a 10-lane bridge 55 meters wide, not a 2-track rail bridge 10 meters wide. Without any tunnels on the route, New Haven-Kingston should cost no more than about $3-3.5 billion in 2020 terms.
Elsewhere, there are small curve easements, even on generally straight portions like in New Jersey and South County, Rhode Island, both of which have curves that if you zoom in close enough and play with the Google Earth circle tool you’ll see are much tighter than 4 km in radius. For the most part this just means building the required structure, and then connecting the tracks to the new rather than old curve in a night’s heavy work; more complex movements of track have been done in Japan on commuter railroads, in a more constrained environment.
There’s a fair amount of taking required. The most difficult segment is New Rochelle-New Haven, with the most takings in Darien and the only tunneling in Bridgeport; the only other new tunnel required is in Baltimore, where it should follow the old Great Circle Tunnel proposal’s scope, not the four-track double-stack mechanically ventilated bundle the project turned into. The Baltimore tunnel was estimated at $750 million in 2008, maybe $1 billion today, and that’s high for a tunnel without stations – it’s almost as high per kilometer as Second Avenue Subway without stations. Bridgeport requires about 4 km of tunnel with a short water crossing, so figure $1-1.5 billion today even taking the underwater penalty and the insane unit costs of the New York region as a given.
A few other smaller deviations from the mainline are worth doing at-grade or elevated: a cutoff in Maryland near the Delaware border in the middle of what could be prime 360 km/h territory, a cutoff in Port Chester and Greenwich bypassing the worst curve on the Northeast Corridor outside major cities, the aforementioned takings-heavy segment through Darien continuing along I-95 in Norwalk and Westport, a short bypass of curves around Fairfield Station. These should cost a few hundred million dollars each, though the Darien-Westport bypass, about 15 km long, could go over $1 billion.
Finally, the variable-tension catenary south of New York needs to be replaced with constant-tension catenary. A small portion of the line, between New Brunswick and Trenton, is being so replaced at elevated cost. I don’t know why the cost is so high – constant-tension catenary is standard around the world and costs $1.5-2.5 million per km in countries other than the US, Canada, and the UK. The Northeast Corridor is four-track and my other examples are two-track, but then my other examples also include transformers and not just wires; in New Zealand, the cost of wires alone was around $800,000 per km. Even taking inflation and four tracks into account, this should be maybe $700 million between New York and Washington, working overnight to avoid disturbing daytime traffic.
The overall cost should be around $15 billion, with rolling stock and overheads. Higher costs reflect unnecessary scope, such as extra regional rail capacity in New York, four-tracking the entire Providence Line instead of building strategic overtakes and scheduling trains intelligently, the aforementioned four-track version of the Baltimore tunnel, etc.
The implications of cheap high-speed rail
I wrote about high-speed rail ridership in the context of Metcalfe’s law, making the point that once one line exists, extensions are very high-value as a short construction segment generates longer and more profitable trips. The cost estimate I gave for the Northeast Corridor is $13 billion, the difference with $15 billion being rolling stock, which in that post I bundled into operating costs. With that estimate, the line profits $1.7 billion a year, a 13% financial return. This incentivizes building more lines to take advantage of network effects: New Haven-Springfield, Philadelphia-Pittsburgh, Washington-Virginia-North Carolina-Atlanta, New York-Upstate.
The problem: building extensions does require the infrastructure on the Northeast Corridor that I don’t think should be in the initial scope. Boston-Washington is good for around a 16-car train every 15 minutes all day, which is very intense by global standards but can still fit in the existing infrastructure where it is two-track. Even 10-minute service can sometimes fit on two tracks, for example having some high-speed trains stop at Trenton to cannibalize commuter rail traffic – but not always. Boston-Providence every 10 minutes requires extensive four-tracking, at least from Attleboro to beyond Sharon in addition to an overtake from Route 128 to Readville, the latter needed also for 15-minute service.
More fundamentally, once high-speed rail traffic grows beyond about 6 trains per hour, the value of a dedicated path through New York grows. This is not a cheap path – it means another Hudson tunnel, and a connection east to bypass the curves of the Hell Gate Bridge, which means 8 km of tunnel east and northeast of Penn Station and another 2 km above-ground around Randall’s Island, in addition to 5 km from Penn Station west across the river. The upshot is that this connection saves trains 3 minutes, and by freeing trains completely from regional rail traffic with four-tracking in the Bronx, it also permits using the lower 4% schedule pad, saving another 1 minute in the process.
If the United States is willing to spend close to $100 billion high-speed rail on the Northeast Corridor – it isn’t, but something like $40-50 billion may actually pass some congressional stimulus – then it should spend $15 billion and then use the other $85 billion for other stuff. This include high-speed tie-ins as detailed above, as well as low-speed regional lines in the Northeast: new Hudson tunnels for regional traffic, the North-South Rail Link, RegionalBahn-grade links around Providence and other secondary cities, completion of electrification everywhere a Northeastern passenger train runs
I hate the term “incremental” when it comes to infrastructure, not because it’s inherently bad, but because do-nothing politicians (e.g. just about every American elected official) use it as an excuse to implement quarter-measures, spending money without having to show anything for it.
So for the purpose of this post, “incremental” means “start with $15 billion to get Boston-Washington down to 3:20 and only later spend the rest.” It doesn’t mean “spend $2 billion on replacing a bridge that doesn’t really need replacement.”
With that in mind, the capacity increases required to get from bare Northeast Corridor high-speed rail to a more expansive system can all be done later. The overtakes on Baltimore-Washington would get filled in to form four continuous tracks all the way, the ones on Boston-Providence would be extended as outlined above, the bypasses on New York-New Haven would get linked to new tracks in the existing right-of-way where needed, the four-track narrows between Newark and Elizabeth would be expanded to six in an already existing right-of-way. Elizabeth Station has four tracks but the only building in the way of expanding it to six is a parking garage that needs to be removed anyway to ease the S-curve to the south of the platforms.
However, one capacity increase is difficult to retrofit: new tracks through New York. The most natural way to organize Penn Station is as a three-line system, with Line 1 carrying the existing Hudson tunnel and the southern East River tunnels, including high-speed traffic; Line 2 using new tunnels and a Grand Central link; and Line 3 using a realigned Empire Connection and the northern East River tunnels. The station is already centered on 32nd Street extending a block each way; existing tunnels going east go under 33rd and 32nd, and all plans for new tunnels continuing east to Grand Central or across the East River go under 31st.
But if it’s a 3-line system and high-speed trains need dedicated tracks, then regional trains don’t get to use the Hell Gate Line. (They don’t today, but the state is spending very large sums of money on changing this.) Given the expansion in regional service from the kind of spending that would justify so much extra intercity rail, a 4-line system may be needed. This is feasible, but not if Penn Station is remodeled for 3 lines; finding new space for a fourth tunnel is problematic to say the least.
The point of integrated timetable planning is to figure out what timetable one want to run in the future and then building the requisite infrastructure. Thus, in the 1990s Switzerland built the tunnels and extra tracks for the connections planned in Bahn 2000, and right now it’s doing the same for the next generation. This can work incrementally, but only if one knows all the phases in advance. If timetable plans radically change, for example because the politicians make big changes overruling the civil service to remind the public that they exist, then this system does not work.
If the United States remains uninterested in high-speed rail, then it’s fine to go ahead with a bare-bones $15 billion system. It’s good, it would generate good profits for Amtrak, it would also help somewhat with regional-intercity rail connectivity. Much of the rest of the system can be grafted on top without big changes.
But then it comes to Penn Station. It’s frustrating, because anything that brings it into focus attracts architects and architecture critics who think function should follow form. But it’s really important to make decisions soon, get to work demolishing the above-ground structures starting when the Madison Square Garden lease runs out, and move the tracks in the now-exposed stations as needed based on the design timetable.
As with everything else, it’s possible not to do it – to do one design and then change to another – but it costs extra, to the tune of multiple billions in unnecessary station reconstruction. If the point is to build high-speed rail cost-effectively, spending the same budget on more infrastructure instead of on a few gold-plated items, then this is not acceptable. Prior planning of how much service is intended is critical if costs are to stay down.
I refined my train performance calculator to automatically compute trip times from speed zones. Open it in Python 3 IDLE and play with the functions for speed zones – so far it can’t input stations, only speed zones on running track, with stations assumed at the beginning and end of the line.
I’ve applied this to a Northeast Corridor alignment between New York and Boston. The technical trip times based on the code and the alignment I drew are 0:36:21 New York-New Haven, 0:34:17 New Haven-Providence, 0:20:40 Providence-Boston; with 1-minute dwell times, this is 1:33 New York-Boston, rising to maybe 1:40 with schedule contingency. This is noticeably longer than I got in previous attempts to draw alignments, where I had around 1:28 without pad or 1:35 with; the difference is mainly in New York State, where I am less aggressive about rebuilding entire curves than I was before.
I’m not uploading this alignment yet because I want to fiddle with some 10 meter-scale questions. The most difficult part of this is between New Rochelle and New Haven. Demolitions of high-price residential properties are unavoidable, especially in Darien, where there is no alternative to carving a new right-of-way through Noroton Heights.
The importance of speeding up the slowest segments
The above trip times are computed based on the assumption that trains depart Penn Station at 60 km/h as they go through the interlocking, and then speed up to 160 km/h across the East River, using the aerodynamic noses designed for 360 km/h to achieve medium speed through tunnels with very little free air. This require redoing the switches at the interlocking; this is fine, switches in the United States are literally 19th-century technology, and upgrading them to Germany’s 1925 technology would create extra speed on the slowest segment.
Another important place to speed up is Shell Interlocking. The current version of the alignment shaves it completely, demolishing some low-rise commercial property in the process, to allow for 220 km/h speeds through the city. Grade separation is obligatory – the interlocking today is at-grade, which imposes unreasonable dependency between northbound and southbound schedules on a busy commuter railroad (about 20 Metro-North trains per hour in the peak direction).
In general, bypasses west of New Haven prioritize the slowest segments of the Northeast Corridor: the curves around the New York/Connecticut state line, Darien, Bridgeport. East of New Haven the entire line should be bypassed until Kingston, even the somewhat less curvy segment between East Haven and Old Saybrook, just because it’s a relatively easy segment where the railroad can mostly twin with I-95 and not have any complex viaducts.
The maximum speed is set at 360 km/h, but even though trains can cruise at such speed on two segments totaling 130 km, the difference in trip time with 300 km/h is only about 3 minutes. Similarly, in southwestern Connecticut, the maximum speed on parts of the line, mostly bypasses, is 250 km/h, and if trains could run at 280 km/h on those segments, which isn’t even always possible given curvature, it would save just 1 minute. The big savings come from turning a 10 miles per hour interlocking into a modern 60 km/h (or, ideally, 90+ km/h) one, eliminating the blanket 120 km/h speed limit between the NY/CT state line and New Haven, and speeding up throats around intermediate stations.
Bypasses are easier to draw than curve modifications. Curves on the Northeast Corridor don’t always have consistent radii – for example, the curves flanking Pawtucket look like they have radius 600 meters, but no, they have a few radii of which the tightest are about 400 meters, constraining speed further. Modifying such curves mostly within right-of-way should be a priority.
Going outside the right-of-way is also plausible, at a few locations. The area just west of Green’s Farms is a good candidate; so is Boston Switch, a tight curve somewhat northeast of Pawtucket whose inside is mostly water. A few more speculative places could get some noticeable trip time improvements, especially in the Bronx, but the benefit-cost ratio is unlikely to be good.
Bush consulting on takings
In some situations, there’s a choice of which route to take – for example, which side of I-95 to go on east of New Haven (my alignment mostly stays on the north side). Some right-of-way deviations from I-95 offer additional choice about what to demolish in the way.
In that case, it’s useful to look for less valuable commercial properties, and try to avoid extensive residential takings if it’s possible (and often it isn’t). This leads to some bush consulting estimates of how valuable a strip mall or hotel or bank branch is. It’s especially valuable when there are many options, because then it’s harder for one holdout to demand unreasonable compensation or make political threats – the railroad can go around them and pay slightly more for an easier takings process.
How fast should trains run?
Swiss planners run trains as fast as necessary, not as fast as possible. This plan does the opposite, first in order to establish a baseline for what can be done on a significant but not insane budget, and second because the expected frequency is high enough that hourly knots are not really feasible.
At most, some local high-speed trains could be designated as knot trains, reaching major stations on the hour or half-hour for regional train connections to inland cities. For example, such a local train could do New York-Boston in 2 hours rather than 1:40, with such additional stops as New Rochelle, Stamford, New London (at I-95, slightly north of the current stop), and Route 128 or Back Bay.
But for the most part, the regional rail connections are minor. New York and Boston are both huge cities, so a train that connects them in 1:40 is mostly an end-to-end train, beefed up by onward connections to Philadelphia, Baltimore, and Washington. Intermediate stops at New Haven and Providence supply some ridership too, much more so than any outlying regional connections like Danbury and Westerly, first because those outlying regional connections are much smaller towns and second much of the trip to those towns is at low speed so the trip time is not as convenient as on an all-high-speed route.
This does not mean Swiss planning maxims can be abandoned. Internal traffic in New England, or in Pennsylvania and South Jersey, or other such regions outside the immediate suburbs of big cities, must hew to these principles. Even big-city regional trains often have tails where half-hourly frequency is all that is justified. However, the high-speed line between Boston and New York (and Washington) specifically should run fast and rely on trips between the big cities to fill trains.
How much does it cost?
My estimate remains unchanged – maybe $7 billion in infrastructure costs, closer to $9-10 billion with rolling stock. Only one tunnel is included, under Bridgeport; everywhere else I’ve made an effort to use viaducts and commercial takings to avoid tunneling to limit costs. The 120 km of greenfield track between New Haven and Kingston include three major viaducts, crossing the Quinnipiac, Connecticut, and Thames; otherwise there are barely any environmentally or topographically sensitive areas and not many areas with delicate balance of eminent domain versus civil infrastructure.
I repeat, in case it is somehow unclear: for $7 billion in infrastructure investment, maybe $8 billion in year-of-expenditure dollars deflated to the early 2020s rather than early 2010s, trains could connect New York and Boston in 1:40. A similar project producing similar trip times between New York and Washington should cost less, my guess is around $3 billion, consisting mostly of resurrecting the old two-track B&P replacement in lieu of the current scope creep hell, building a few at-grade bypasses in Delaware and Maryland, and replacing the variable-tension catenary with constant-tension catenary.
None of this has to be expensive. Other parts of the world profitably build high-speed rail between cities of which the largest is about the size of Boston or Philadelphia rather than the smallest; Sweden is seriously thinking about high-speed trains between cities all of which combined still have fewer people than metropolitan Boston. Better things are possible, on a budget, and not just in theory – it’s demonstrated every few years when a new high-speed rail line opens in a medium-size European or Asian country.
After drawing a map of an integrated timed transfer intercity rail network for the state of New York, people asked me to do other parts of the United States. Here is New England, with trains running every 30 minutes between major cities:
New England is a much friendlier environment for intercity rail growth than Upstate New York, but planning there is much more delicate. The map thus has unavoidable omissions and judgment calls, unlike the New York map, which straightforwardly follows the rule of depicting intercity lines but not suburban lines like the Long Island network. I ask that people not flame me about why I included X but not Y without reading the following explanations.
The tension between S-Bahn and ITT planning
The S-Bahn concept involves interlining suburban rail lines through city center to provide a high-frequency urban trunk line. For example, trains from a number of East Berlin neighborhoods and Brandenburg suburbs interline to form the Stadtbahn: in the suburbs, they run every 10 or 20 minutes, but within the Ring, they combine to form a diameter running regularly every 3:20 minutes.
The integrated transfer timetable concept instead involves connecting different nodes at regular intervals, typically half an hour or an hour, such that trains arrive at every node just before a common time and leave just after, to allow people to transfer. In a number of major Swiss cities, intercity trains arrive a few minutes before the hour every 30 minutes and depart a few minutes after, so that people can connect in a short amount of time.
S-Bahn and ITT planning are both crucial tools for good rail service, but they conflict in major cities. The ITT requires all trains to arrive in a city around the same time, and depart a few minutes later. This forces trains from different cities to have different approach tracks; if they share a trunk, they can still arrive spaced 2-3 minutes apart, but this lengthens the transfer window. The idea of an S-Bahn trunk involves trains serving the trunk evenly, which is not how one runs an ITT.
Normally, this is no problem – ITTs are for intercity trains, S-Bahns are for local service. But this becomes a problem if a city is so big that its S-Bahn network grows to encompass nearby city centers. In New York, the city is so big that its shadow reaches as far as Eastern Long Island, New Haven, Poughkeepsie, and Trenton. Boston is smaller but still casts shadows as far as southern New Hampshire and Cape Cod.
This is why I don’t depict anything on Long Island on my map: it has to be treated as the extension of an S-Bahn system, and cannot be the priority for any intercity ITT. This is also true of Danbury and Waterbury: both are excellent outer ends for an electrified half-hourly regional rail system, but setting up the timed transfers with the New Haven Line (which should be running every 10 minutes) and with high-speed rail (which has no reason to stop at the branch points with either Danbury or Waterbury) is infeasible. In Boston I do depict some lines – see below on the complications of the North-South Rail Link.
The issue of NSRL
The North-South Rail Link is a proposed north-south regional rail tunnel connecting Boston’s North and South Stations. Current plans call for a four-track tunnel extending across the river just north of North Station, about 4.5 km of route; it should cost $4 billion including stations, but Massachusetts is so intent on not building it lies that the cost is $12 billion in 2018 dollars.
In common American fashion, NSRL plans are vague about how service is to run through the tunnel. There are some promises of running intercity trains in addition to regional ones; Amtrak has expressed some interest in running trains through from the Northeast Corridor up to the northern suburbs and thence to Maine. However, we are not engaging in bad American planning for the purposes of this post, but in good Central European planning, and thus we must talk about what trains should run and design the tunnel appropriately.
The rub is that Boston’s location makes NSRL great for local traffic and terrible for intercity traffic. When it comes to local traffic, Boston is right in the middle of its metropolitan region, just offset to the east because of the coast. The populations of the North Side and South Side suburbs are fairly close, as are their commuter volumes into Boston. Current commuter rail ridership is about twice as high on the South Side, but that’s because South Station’s location is more central than North Station’s. NSRL really is a perfect S-Bahn trunk tunnel.
But when it comes to intercity traffic, Boston is in the northeast corner of the United States. There are no major cities north of Boston – the largest such city, Portland, is a metro area of 600,000. In contrast, going south, New York should not be much more than an hour and a half away by high-speed rail. Thus, high-speed rail has no business running through north of Boston – the demand mismatch south and north is too high.
Since NSRL is greatly useful for regional traffic but not intercity traffic, the physical infrastructure should be based on S-Bahn and not ITT principles, even though the regional network connects cities quite far away. For one, the tunnel should require all trains to make all stops (South Station, Aquarium, North Station) for maximum local connectivity. High-speed trains can keep feeding South Station on the surface, while all other traffic uses the tunnel.
But on the North Side, feeding North Station on the surface is not a good idea for intercity trains. The station is still awkwardly just outside city center. It also offers no opportunity to transfer to intercity trains to the most important city of all, New York.
The only resolution is to treat trains to Portland and New Hampshire as regional trains that just go farther than normal. The Nashua-Manchester-Concord corridor is already as economically linked to Boston as Providence and Worcester, and there are plans for commuter rail service there already, which were delayed due to political opposition to spending money on trains from New Hampshire Republicans after their 2010 election victory. Portland is more speculative, but electric trains could connect it with Boston in around an hour and a half to two hours. These trains would be making suburban stops north of Boston that an intercity train shouldn’t normally make, but it’s fine, the Lowell Line has wide stop spacing and the intermediate stops are pretty important post-industrial cities. At Portland, passengers can make a timed connection to trains to Bangor, on the same schedule but with shorter trainsets as the demand north of Portland is much weaker.
On the map, I also depict Boston-Cape Cod trains, which like Boston-Concord trains are really suburban trains but going farther. Potentially, the branch to Cape Cod – the Middleborough branch of the Old Colony Lines – could even run through with the Lowell Line, either the branch to Concord or the Wildcat Branch to Haverhill and Portland. Moreover, the sequencing of the branches should aim to give short connections to Boston-Albany high-speed trains as far as reasonable.
The issue of the Northeast Corridor
The Northeast Corridor wrecks the ITT plan in two ways, one substantial and one graphical.
The snag is that there should be service on legacy track running at a maximum speed of 160-200 km/h in addition to high-speed service on high-speed tracks. There may be some track sharing between New York and New Haven to reduce construction costs, using timed overtakes instead of full track segregation, but east of New Haven the high-speed trains should run on a new line near I-95 to bypass the Shore Line’s curves, and the Shore Line should be running electric regional trains to connect to the intermediate cities.
The graphical problem is that the distance between where the legacy route is and where the high-speed tracks should be is short, especially west of New Haven, and depicting a red line and a blue line together on the map is not easy. I will eventually post something at much higher resolution than 1 pixel = 500 meters. This also affects long-distance regional lines that I’d like to depict on the map but connect only to legacy trains on the Northeast Corridor, that is the Danbury and Waterbury Branches.
For planning purposes, figure that both run every half hour all day, are electric, run through to and beyond New York as branches of the New Haven Line, and are timed to have reasonable connections to high-speed trains to Albany and points north in New York. Figure the same for trains between New Haven and Providence, with some additional runs in the Providence suburbs giving 15-minute urban frequencies to such destinations as Olneyville and Cranston.
The substantial issue is that the Northeast Corridor is far too high-demand for a half-hourly ITT. Intercity trains run between New York and Boston better than hourly today, and that’s taking twice as long as a TGV and charging 2.5-4 times as much. My unspoken assumption when planning how everything should fit together is that there should be a 400-meter long train every 15 minutes on the corridor past New Haven, spaced evenly around Boston to overtake regional trains to Providence at consistent locations. Potentially, there should be more local trains taking around 1:50 and more express trains taking around 1:35, and then all timed transfers should be to the local trains.
On the New Haven Line, too, regional rail demand is much more than a train every half hour. Trains run mostly every half hour today, with management that is flagrantly indifferent to off-peak service, and trip times that are about 50% longer than they should be. Nonetheless, best practice is to set up timed transfers such that various branches all connect to the same train, so that passengers can connect between different branches. This mostly affects Waterbury; it’s useful to ensure that Waterbury trains arrive at Bridgeport with a short transfer to a train toward New Haven that offers a quick connection to trains to points north and east.
Planning HSR around timed connections
Not counting lines that are in the Boston sphere, or the lines around Albany, which I discussed two weeks ago, there are three lines proposed for timed connection to high-speed rail: New London-Norwich, Providence-Worcester-Fitchburg, Springfield-Greenfield.
All three are regional lines, not intercity lines. They are not optimized for intercity speed, but instead make a number of local urban and suburban stops. This is especially true of Springfield-Northampton-Greenfield, a line that Sandy Johnston and I have been talking about since 2014. A Springfield-Greenfield line with 1-2 intermediate stops might be able to do a one-way trip in around 39 minutes, at which point a 45-minute operator schedule may be feasible with a very tight turnaround regime – but there’s enough urban demand along the southern half of the route that adding stops to make it about 50 minutes with a one-hour operator schedule is better.
The Providence-Worcester line is likewise slower than it could be if it were just about Providence and Worcester. The reason is that high-speed rail compresses distances along its route. Providence-Boston by high-speed rail is about 22 minutes nonstop, including schedule contingency. Boston-Worcester is about the same – slower near Boston because of scheduling difficulties along the Turnpike and the inner Worcester Line, faster near the outer end because Worcester has no chance of getting a city center station but rather gets a highway station. Now, passengers have a range of transfer penalties, and to those who are averse to connections and have a high personal penalty, the trip between the two cities is more attractive directly than via Boston. But there are enough passengers who’d make the trip via Boston that the relative importance of intermediate points grows: Pawtucket, Woonsocket, Uxbridge, Millbury. In that situation, the importance of frequency grows (half-hourly is a must, not hourly) and that of raw speed diminishes.
The onward connection to Fitchburg is about three things. First, connecting Providence with Fitchburg. Second, connecting Worcester with Fitchburg. And third, connecting Fitchburg with the high-speed line. This makes investments into higher speed more valuable, since Fitchburg’s importance is high compared with that of points between Worcester and Fitchburg. The transfer between the line and high-speed rail should be timed in the direction of Fitchburg-to-Albany first of all, and Providence-to-Albany second of all, as the connections from the endpoints to Boston are slower than direct commuter trains.
The presence of this connection also forces the Worcester station to be at the intersection with the line to Providence. Without this connection, it may be better to site the station slightly to the west, at 290 rather than 146, as the area already has Auburn Mall.
Finally, the New London-Norwich line is a pure last-mile connector from the New London train station, which is forced to be right underneath the I-95 bridge over the river, to destinations to the north. The northern anchor is Norwich opposite the historic center, but the main destination is probably the Mohegan Sun casino complex. Already there are many buses connecting passengers from New York to the casino. The one-way trip time should be on the order of 21-22 minutes, but with a turnaround it’s a 30-minute schedule, and the extension south to the historic center of New London is for completeness; with a timed connection, trains could get between Penn Station and Norwich in around 1:20 counting connection time, and between Penn Station and Mohegan Sun in maybe 5 minutes less.
What about Vermont?
Vermont’s situation is awkward. Burlington is too far north and too small to justify a connection to high-speed rail by itself. A low-speed connection might work, but the line from Burlington south points toward Rutland and not New York, and connecting it onward requires reversing direction. If Vermont had twice its actual population this might be viable, but it doesn’t.
But Vermont is right between New York and Montreal. I generally don’t show New York-Montreal high-speed rail on my maps. It’s a viable line, but people in both cities severely overrate it, especially compared with New York-Toronto; I have to remind readers this whenever I write about international high-speed trains. In the event such a line does open, Burlington is the only plausible location for a Vermont stop – everything else is too small, even towns that historically did have rail service, like Middlebury. Rutland could get a line running partly on high-speed track and partly on legacy track taking it down to Glens Falls or Saratoga Springs to transfer to onward destinations, or maybe Albany if trains run 2-3 minutes apart in pairs every 30 minutes.
Current plans for Vermont try to connect it directly to Boston via New Hampshire, and that is wrong. The Vermonter route is mountainous from Greenfield to Burlington; trains will never be competitive with driving there. Another route under occasional study going into Boston from the north was even included on a 2009 wishlist of high-speed rail routes, under the traditional American definition of high-speed rail as “train that is faster than a sports bicycle.” That route, crossing mountains in both New Hampshire and Vermont, is even worse. The north-south orientation of the mountains in both states forces east-west routes to either stick to the lowlands or consolidate to strong enough routes that high-speed rail tunnels are worthwhile.
How much does this cost?
As always, I am going to completely omit the Northeast Corridor from this cost analysis; an analysis of that will happen later, and suffice is to say, the benefit-cost ratio if there’s even semi-decent cost control is extremely high.
With that in mind, the central pieces of this program are high-speed lines from Boston to Albany and from New Haven to Springfield, in a T system. The 99 km New Haven-Springfield line, timetabled at 45 minutes including turnaround and maybe 36 minutes in motion, is on the slow side for high-speed rail, as it is short and has a crucial intermediate station in Hartford. It does not need any tunnels or complex viaducts, and property takings are nonzero but light; the cost should not be higher than about $2-2.5 billion, utilizing legacy track for much of the way.
The Boston-Albany line is much costlier. It’s 260 km, and crosses the aforementioned north-south mountains in Western Massachusetts. Tunnels are unavoidable, including a few kilometers of digging required just west of Springfield to avoid a slowdown on suburban curves. At the Boston end, tunneling may also be unavoidable next to the Turnpike. The alternative is sharing a two-track narrows with the MBTA Worcester Line in Newton; it’s possible if the trains run no more than every 15 minutes, which is a reasonable short-term imposition but may be too onerous in the longer term if better service builds up more demand for commuter rail frequency in Newton. My best guess is that without Newton, the line needs around 20 km of tunnel and can piggyback on 35 km of existing lines at both ends, for a total cost in the $6-8 billion range. This figure is sensitive to whether my 20 km estimate is correct, but not too sensitive – at 40 it grows to maybe $9 billion, at 0 it shrinks to $4.5 billion.
Estimating the costs of the blue lines on the map is harder. All of them are, by the standard of high-speed rail, very cheap per kilometer. A track renewal machine on a one-third-in-tunnel German high-speed line can do track rebuilding for about a million euros per single-track-kilometer. All of these lines would also need to be electrified from scratch, for $1.5-3 million per kilometer. Stations would need to be built, for a few million apiece. My first-order estimate is $1 billion for the three blue connector lines and about the same for Boston-Portland-Bangor; the Hyannis and Concord lines would go in a regional rail basket. The NSRL tunnel should be $4 billion or not much more, and not what Massachusetts wants voters to believe it is to justify its decision not to build it.
The reason for the relatively limited map (e.g. no Montreal service) is that these lines are not such slam dunks that they’re worth it at any price. Cost control is paramount, subject to the bare minimum of good service (e.g. electrification and level boarding). For what I think a fair cost is, those lines are still good, providing fast connectivity across New England from most places to most other places. Moreover, the locations of the major nodes, like Worcester and Springfield, allow timing bus interchanges as well, providing further connections to various suburbs and city neighborhoods.
The red high-speed lines are flashy, but the blue ones are important too. That’s the key takeaway from planning in Switzerland, Austria, and the Netherlands, all of which have high rail usage without great geography for intercity rail. Trains should be planned coherently as a network, with all parts designed in tandem to maximize connectivity. This isn’t just about going between Boston and Springfield or Boston and Albany or New Haven and Springfield, but also the long tail of weaker markets using timed connections, like New Haven-Amherst, Brockton-Worcester, Dover-Providence, Stamford-Mohegan Sun, and so on. A robust rail network based on ITT design principles could make all of these and many more connections at reasonable cost and speed.
At a meeting with other TransitMatters people, I had to explain various distinctions in what is called in American parlance regional rail or commuter rail. A few months ago I wrote about the distinction between S-Bahn and RegionalBahn, but made it clear that this distinction was about two different things: S-Bahns are shorter-distance and more urban than RegionalBahns, but they’re also more about service in a contiguous built-up area whereas RegionalBahns have the characteristics of interregional service. In this post I’d like to explore the different travel markets for regional rail not as a single spectrum between urban and long-range service, but rather as two distinct factors, one about urbanity or distance and one about whether the line connects independent centers (“interregional”) or a monocentric urban blob (“intraregional”).
This distinction represents a two-dimensional spectrum, but for simplicity, let’s start with a 2*2 table, so ubiquitous from the world of consulting:
|Connection \ Range||Short||Long|
|Intraregional||Urban rail, S-Bahn||Big-city suburban rail|
|Interregional||Polycentric regional rail||RegionalBahn|
The notions of mono- and polycentricity are relative. Downtown Providence, Newark, and San Jose all have around 60,000 jobs in 5 km^2. But Caltrain and the Providence Line are both firmly in the RegionalBahn category, the other end being Downtown San Francisco or Boston, 70-80 km away with 300,000-400,000 jobs in 5-6 km^2. Newark, in an essentially contiguous urban area with New York, 16 km from Midtown and its 1.2 million jobs in 6 km^2, is relatively weaker and does not fit into the interregional category; a New York-Newark line is an S-Bahn.
On the 2*2 table, the appellations “big-city” and “polycentric” are necessary. This is because longer-range rail lines are likelier to get out of the city and its immediate suburbs and connect to independent urban centers. Exceptions mostly concern the size of the primary urban cluster. If it is large, like New York, it can cast a shadow for tens of kilometers in each direction: commuter volumes are high from deep into Long Island, as far up the Northeast Corridor as Westport, as far up the Hudson as northern Westchester, and so on. In Paris, I wouldn’t be comfortable describing any of the RER and Transilien lines as RegionalBahn. In London, the closest independent cities of reasonable size are Cambridge, Brighton, Oxford, and Portsmouth, the first two about 80 km away and the last two about 100.
Tokyo, about as big as New York and London combined, casts an even longer shadow. In my post on S-Bahns and RegionalBahns I called some of its outer regional rail branches RegionalBahn, giving the examples like the Chuo Line past Tachikawa. But even that line is not really interregional in any meaningful way. It stays within the Tokyo prefecture as far as Takao, 53 km from Tokyo Station, and commuter service continues until Otsuki at kp 88, but everything along the line is bedroom communities for Tokyo or outright rural. The branching and short-turns at Tachikawa mean that the Chuo Line through Tachikawa is a long S-Bahn, and past Tachikawa is really a suburban commuter line too long to be an S-Bahn but too monocentric and peaky to be Regionalbahn (the peak-to-base frequency ratio is about 2:1, whereas German RegionalBahn is more commonly 1:1).
At the other end, we can have regional rail that is short-range but connects two distinct centers. This occurs when relatively small cities are in proximity to each other. In a modern first-world economy, these cities would form a polycentric region, like the Rhine-Ruhr or Randstad. Smaller regions with these characteristics include the Research Triangle, where relatively equal-size Raleigh and Durham are 40 rail kilometers apart, and Nord, where Lille is 30-50 km from cities like Douai and Valenciennes. This may even occur in a region with a strong primary center, if the secondary center is strong enough, as is the case for Winterthur, 28 km from Zurich, which has Switzerland’s fourth highest rail ridership.
Size is measured in kilometers, not people. Stockholm is a medium-size city region, but Stockholm-Uppsala is firmly within RegionalBahn territory, as the two cities are 66 km apart. Randstad’s major cities are all closer to each other – Amsterdam-Rotterdam is about 60 km – and that’s a region of 8 million, not 3 million like Stockholm and the remainder of Uppland and Södermanland.
The issue of frequency
The importance of the 2*2 table is that distance and urban contiguity have opposite effects on frequency: high frequency is more important on short lines than on long lines, and matching off-peak frequency to peak frequency is more important on interregional than intraregional lines.
Jarrett Walker likes to say that frequency is freedom, but what frequency counts as freedom depends on how long passengers are expected to travel on the line. Frequency matters insofar as it affects door-to-door travel time including wait time, so it really ought to be measured as a fraction of in-vehicle travel time rather than as an absolute number. An urban bus with an average passenger trip time of 15 minutes should run every 5 minutes or not much longer; if it runs every half hour, it might as well not exist, unless it exists for timed connections to longer-range destinations. But an intercity rail line where major cities are 2 hours apart can easily run every half hour or even every hour.
The effect of regional contiguity is more subtle. The issue here is that an intraregional line is likely to be used mostly by commuters at the less dense end. The effect of distance can obscure this, but within a large urban area, a 45-minute train will be full of commuters traveling to the primary city in the morning and back to the suburbs in the afternoon or evening; the same train between two distinct cities, like Boston and Providence, will not have so many commuters. In contrast, the same 45-minute trip will get much more reverse-commute travel and slightly more non-commute travel if it connects two distinct cities, because the secondary city is likelier to have destinations that attract travelers.
In no case are the extreme peak-to-base ratios of American commuter lines justifiable. Lines with tidal commuter flows can run 2:1 peak-to-base ratios, as is common in Tokyo, but much larger ratios waste capacity. The marginal cost of service between the morning and afternoon peaks is so low until it matches peak service that having less midday than peak service at all is only justifiable in very peaky environments. The 45-minute suburbs of New York, Tokyo, and other huge cities can all live with a 2:1 ratio, but other lines should have lower ratios, and interregional lines should have a 1:1 ratio.
The implication is that just as peak-to-base ratios going as high as 2:1 are acceptable for long-range intraregional lines, short-range interregional lines must run a constant, high frequency all day. I would groan at the thought of even half-hourly frequency on a 40-km interregional line; the worst I’m comfortable with is 15-20 minutes all day. Of note, such lines are necessarily pretty fast, since by assumption they make few intermediate stops to speed up travel between the two main cities – if there are significant cities in the middle then the lines connect even shorter-range cities and should be even more frequent.
Urban, suburban, intercity
Individual lines may have the characteristics of multiple variants of regional rail. They pass through urban neighborhoods on their way to outlying areas, which may be suburbs or independent cities; they may also pass through multiple kinds of independent areas.
In practice, in big cities this leads to three tiers on the same line: urban at the inner end, suburban at the middle end, interregional at the outer end. Inversions, in which there are independent cities and then suburbs, are possible but extremely rare – I can’t think of any in Paris, London, or New York, and arguably only three in Tokyo (Chiba, Saitama, Yokohama); fundamentally, if there are suburbs of the primary city beyond your municipality, then your municipality is likely to itself be popular as a suburb of the primary city.
That regional lines have these three tiers of demand type does not mean that every single regional line does. Some lines don’t reach any significant independent city. Some don’t usefully serve close-in urban areas – for example, the Providence Line barely serves anything urban, since the stop spacing is wide in order to speed up travel to high-demand suburbs and to Providence and the closest-in urban neighborhoods have Orange Line subway service. In rare cases, the suburban tier may be skipped, because there just isn’t much tidal suburban commuter ridership; in Boston, the Newburyport Line is an example, since its inner area has unbroken working-class urban development almost all the way to Salem, and then there’s almost nothing between Salem and Newburyport.
This does not mean that suburbs are always in between urban areas and independent cities – this is just a specific feature of large metropolitan areas. In smaller ones, the middle tier between urban and long-range interregional service is occupied by short-range interregional service rather than suburban commuter rail. Skipping the suburban tier, which is rare enough in large cities that in the cities I think about most often the only example I can come up with is the Newburyport Line, is thus completely normal in smaller cities.
There are common best practices for commuter rail: electrification, level boarding, frequent clockface schedules, timed transfers, fare integration, proof of payment fare collection.
However, high frequency means different things on lines of different characteristics. An interregional line should be running consistent all-day frequency, and if it is long enough could make do with half-hourly trains with timed connections to suburban buses; an urban line should be running every few minutes as if it were a metro line. Regional rail lines with characteristics off the main diagonal of the S-Bahn to RegionalBahn spectrum have different needs – suburban lines can have high peak frequency to reduce road congestion, although they should still have useful off-peak frequency; short-range interregional lines should run every 10-20 minutes all day.
The distinctions between intraregional and interregional lines and between short- and long-range lines may also affect other aspects of planning: station spacing, connections to local surface transit, connections at the city center end, through-running, etc. Even when the best industry practices are the same in all cases, the relative importance of different aspects may change, which changes what is worth spending the most money on.
Since an individual line can serve multiple markets on its way from city center to a faraway outlying terminal, it may be useful to set up a timetable that works for all of these markets and their differing needs. For example, urban lines need higher frequency than suburban and interregional ones, so a regional line with significant urban service should either branch or run short-turn trains to beef up short-range frequency. If there is a suburban area in the middle with demand for high peak frequency but also a secondary city at the outer end, it may be useful to give the entire line high all-day frequency, overserving the line off-peak just because the cost of service is low.
Ultimately, regional rail is about using mainline rail to fulfill multiple functions; understanding how these functions works is critical for good public transportation.
I wrote about how the future is not retro, and Daniel Herriges Strong Towns just responded, saying that traditional development is timeless. I urge all readers to click the last link and read the article, which makes some good points about how cars hollowed out what both Daniel and I call the traditional prewar Midwestern town. There are really two big flaws in the piece. First, it makes some claims about inequality and segregation that are true in American cities but false in the example I give for spiky development, Vancouver. And second, it brings up the resilience of the traditional small town. It’s the second point that I wish to contest: small is not resilient, and moreover, as the economy and society evolve, the minimum size required for resilience rises.
Small cities in the 2010s
In the premodern era, a city of 50,000 was a bustling metropolis. In 1900, it was still a sizable city. In 2019, it is small. The difference is partly relative: a migrant to the big city had the option of moving to a few 200,000 cities in 1900 and one of about ten 1,000,000+ cities, whereas today the same migrant can move to many metro areas with millions of people. But part of it has to do with changes in the economy.
In Adam Smith’s day, big businesses were rare. If you had five employees, you were a big employer. Then came the factory system and firm size grew, but even then companies were small by the standards of today’s specialized economy. A city of 50,000 might well specialize in a single product, as was common in the American manufacturing belt (Krugman mentions this on pp. 11-12 here), but there would be many factories each with a few hundred employees.
But as the economy grows more complex, firm size grows, and so does the interdependence between different firms in the same supply chain. Moreover, the support functions within a city grow in complexity: schools, a hospital, logistics, retail, and so on. The proportion of the population employed in the core factory is lower, as the factory’s high productivity supports more non-manufacturing employees. The upshot is that it’s easy for a town of 50,000 to live off of a single firm and its supply chain. This is not resilient: if the firm fails, the town dies.
Occasionally, cities of that size can have more resilience. Perhaps they’re suburbs of a larger city, in which case they live off of commuting to a more diverse economic center. Perhaps they happen to live off of an industry that cannot die so easily, such as a state capital or a university. On social media one of my followers brought up farming as an example of an activity whose towns have held up in the Midwest better than manufacturing towns; farming is in fact extremely risky, but it has been subsidized since the 1930s, so it has some resilience thanks to subsidies from more internally resilient parts of the country.
Large cities and resilience
I read Ed Glaeser not so much for his observations about the housing market – he’s a lot of things but he’s not a housing economist – as for his economic history. He has a pair of excellent papers describing the economic histories of Boston and New York respectively. Boston, he argues, has reinvented itself three times in the last 200 years after declining, using its high education levels to move up the value chain. New York was never in decline except in the 1970s, and has resiled from its 1980 low as well.
These as well as other large cities have economic diversity that small cities could never hope to have. At the time Glaeser wrote his paper about New York, in 2005, the city seemed dominated by finance and related industries. And yet in the 2007-9 recession, which disproportionately hit finance, the metro area’s per capita income relative to the national average barely budged, falling from 135.3% to 133.8%; in 2017 it was up to 137.5%. The New York region is a center of finance, yes, but it’s also a center of media, academic research, biotech, and increasingly software.
New York is extremely large, and has large clusters in many industries, as do London, Paris, Tokyo, and other megacities. But even medium-size cities often have several clusters, if not so many. This is especially evident in Germany, where Munich, Hamburg, Stuttgart, and Frankfurt are not particularly large. Munich is the center of conglomerates in a variety of industries, including cars (BMW, far and away the largest employer, but also MAN), general industry (Siemens), chemicals (Linde), and finance (Allianz).
What’s true is that these large cities have much more knowledge work than menial work – yes, even Munich, much more a center of engineering than of menial production. But the future is not retro in the mix of jobs any more than it is in its urban layout. The nostalgics of the middle of the 20th century taxed productive industrial cities to subsidize farmers, treating industrial work as the domain of socialists, Jews, immigrants, and other weirdos; the nostalgics of the early 21st century propose to tax productive knowledge economies to subsidize menial workers, and in some specific cases, like American protection of its auto industry, this has been the case for decades.
Small cities as suburbs
In Germany, Switzerland, and the Netherlands, unlike in the United States or France, there is a vigorous tradition of historic small cities becoming suburbs of larger cities while retaining their identity. This doesn’t really involve any of Strong Towns’ bêtes noires about roads and streets – in fact pretty much all of these cities have extensive sprawl with big box retail and near-universal car ownership. Rather, they have tight links with larger urban cores via regional rail networks, and German zoning is less strict about commercialization of near-center residential areas than American zoning. There was also no history of white flight in these areas – the white flight in Germany is in the cores of very large cities, like Berlin, which can replace fleeing whites one to one with immigrants.
In this sense, various Rhineland cities like Worms and Speyer do better than Midwestern cities of the same size. But even though they maintain their historic identities, they are not truly economically independent. In that sense, a better American analogy would be various cities in New England and the mid-Atlantic that have fallen into the megalopolis’s orbit, such as Salem, Worcester, Providence, Worcester, New Brunswick, and Wilmington. Many of these are poor because of the legacy of suburbanization and white flight, but their built-up areas aren’t so poor.
However, the most important link between such small cities and larger urban core, the regional railway, heavily encourages spiky development. In Providence, developers readily build mid-rise housing right next to Providence Station. If the quality of regional rail to Boston improves, they will presumably be willing to build even more, potentially going taller, or slightly farther from the station. Elsewhere in the city, rents are not high enough to justify much new construction, and Downcity is so weak that the tallest building, the Superman Building, is empty. In effect, Providence’s future economic value is as part of the Boston region.
The relatively even development of past generations is of less use in such a city. The economy of a Providence or a Wilmington is not strong enough that everyone can work in the city and earn a good wage. If the most important destination is a distant core like Boston or Philadelphia, then people will seek locations right near the train station. Driving is not by itself useful – why drive an hour from Rhode Island when cheaper suburbs are available within half an hour? Connecting from local transit would be feasible if the interchange were as tightly timed and integrated as in Germany, but even then this system would be oriented around one dot – the train station – rather than a larger walkable downtown area.
A bigger city is a better city
Resilience in the sense of being able to withstand economic shocks requires a measure of economic diversity. This has always been easier in larger cities than in smaller ones. Moreover, over time there is size category creep: the size that would classify a city a hundred years ago as large barely qualifies it to be medium-size today, especially in a large continental superpower like the US. As global economic complexity increases, the size of businesses and their dedicated supply chains as well as local multipliers rises. The city size that was perfectly resilient in an economy with a GDP per capita of $15,000 is fragile in an economy with a GDP per capita of $60,000.
Usually, the absolute richest or more successful places may not be so big. There are hundreds of American metro areas, so a priori there is no reason for New York to be at the top, just as there is no reason for it to be at the bottom. Nonetheless, the fact that larger cities are consistently richer as well as at less risk of decline than smaller cities – New York is one of the richest metro areas, just not the single richest – should give people who think small is beautiful pause.
Whatever one’s aesthetic judgment about the beauty of the upper Mississippi versus that of the lower Hudson, the economic and social system of very large places weathers crises better, and produces more consistent prosperity. Economically and socially, a bigger city is a better city, and national development policy should reject nostalgia and make it possible for developers to build where there is demand – that is, in the richest, most populated metro areas, enabling these regions to grow further by infill as well as accretion. Just as 50,000 was fine in 1900 but isn’t today, a million is fine today but may not be in 2100, and it’s important to enable larger cities to form where people want to live and open businesses.
Remember how ten years ago the American urbanist conversation was all about carving the country up into megaregions? The America 2050 project drew some lines connecting metro areas into regions, designed to imitate the Boston-Washington corridor in concept, and asserted that this would be the future of American growth. The concept seems to have dropped off the discourse, and for good reason, but it may be useful to have a second look. The Boston-Washington megalopolis is a genuine megaregion, and it’s useful to see which regions elsewhere in the world share its characteristics.
The key takeaway is that rich cities do not have to be in megaregions. The Northeast Corridor is a rich megaregion, and San Francisco, Los Angeles, and Chicago anchor smaller megaregions of their own; but in Europe, among the richest cities only Frankfurt and Amsterdam are in megaregions, while London, Paris, Hamburg, and Munich are not. Megaregions are areas of high population density and interlinked social networks. Their size may give them economic advantage, but it doesn’t have to; urbanists and urban geographers must avoid overselling their importance.
What is a megaregion?
The original Boston-Washington megalopolis was defined in the 1960s, as a linear region with continuous suburban sprawl. The core comes from New York and Philadelphia, which share some suburbs in Central Jersey, their regional rails meeting at Trenton. However, continuous sprawl goes north to New Haven, Hartford, and Springfield, with only a few tens of km of separation from Providence and Worcester on the way to Boston; and southwest to Baltimore and Washington, with suburbs spaced closely together along the I-95 corridor.
There are extensive academic connections. Academics are generally hypermobile, but form especially thick metropolitan connections. Living in Boston and reverse-commuting to Brown is normal, and people at Brown would sometimes go up to Harvard or MIT for seminars when sufficiently important or interesting people gave talks. Connections up and down the central part of the corridor are extensive as well, stretching from Yale down to Penn. There is a gap between New Haven and Providence, as Hartford and Springfield aren’t academic centers; perhaps for academics the megaregion only stretches from New Haven to Washington, but even so, at least two-thirds of the megaregion remains intact.
Socially, there are strong connections along the corridor as well. They’re rarely end-to-end, but people in fandom routinely go a state or two over for conventions, so conventions in Connecticut and Rhode Island draw from New York and Boston, conventions in New Jersey draw from Philadelphia and New Haven, and conventions in Maryland draw from Philadelphia and Northern Virginia. On some stretches, weekend trips are normal, like the Columbia students who’d go back to visit parents in suburban Philadelphia every weekend, or people in New York who dated people in New Haven and didn’t even really think of it as a long-distance relationship.
Which regions qualify as megaregions?
Outside the Northeast, it is difficult for me to judge the extent of social connections, with a few key exceptions. However, I can judge how continuous urbanization is and, using American survey data on commuting, whether two adjacent core urban areas share suburbs. In Europe, I do not have commuting data, but it is easy to look at regional rail maps and see when S-Bahn networks touch.
In the United States, the three largest core metropolitan areas outside the Northeast – Los Angeles, Chicago, and San Francisco – all anchor megaregions. However, in all three cases, the big core metro area dominates the broader region. Los Angeles has continuous sprawl down the coast to San Diego, and the two metro areas’ commuter rail networks touch; Chicago similarly has continuous sprawl up to Milwaukee, and if Milwaukee bothered to run regional trains then they would probably go down to Kenosha and connect to Metra; the Bay Area’s high housing costs have driven many people to the San Joaquin Delta, most of the way to Sacramento, and the Amtrak route connecting San Jose and Oakland with Sacramento is largely planned as regional rail nowadays.
New York is of course much larger than the other core regions of the megalopolis, but its metro area has at most half the population of the region, and even that requires making the broadest assumptions on what counts as part of the metro area and the narrowest ones on what counts as part of the megalopolis. If metro New York excludes mostly economically independent areas like New Haven and Central Jersey, and the megalopolis includes some inland areas like Albany and Harrisburg, then New York is only one third of the megalopolis. In contrast, the five-county Los Angeles metro area has three quarters of Southern California’s population, the Bay Area has about two thirds of its megaregion’s population, and metro Chicago has about 85% of the combined population of Chicago and Milwaukee.
Suburb sharing in smaller megaregions
High population density and suburban sprawl can lead some core urban areas to share suburbs, forming a megaregion with much lower population than the megalopolis. Florida supplies at least one such example: out of 237,000 employed residents in Polk County, 26,000 commute to Orlando’s Orange County and 29,000 commute to Tampa’s Hillsborough County and St. Petersburg’s Pinellas County; the western parts of Polk County have a higher density of Tampa-bound commuters and the eastern parts have a higher density of Orlando-bound commuters, but there is a fair amount of mixing, as well as anywhere-to-anywhere commuting within the county. By all accounts, Orlando and Tampa should be placed into one megaregion.
South Florida is arguably a megaregion as well. It is treated as a metro area stretching from Miami or even Key West north to West Palm Beach, but its northern, central, and southern areas have distinct urban cores. Miami-Dade County has 982,000 employed residents, of whom only 28,000 work in Palm Beach County; in the other direction, 29,000 workers from Palm Beach commute to Miami-Dade out of 513,000. This megaregion stretches even further north – St. Lucie County has 13,000 out of 100,000 workers commuting to Palm Beach County – but there is a gap in both population density and commuting zones between Port St. Lucie and Space Coast. Socially, too, the people I know on Space Coast don’t have ties to South Florida, and barely have any to Orlando. So the bulk of Florida is really two linear megaregions, one north-south and one southwest-northeast, which may be close but do not merge.
Finally, crossing the Pond, Northern England features a megaregion out of core metro areas of similar size to those of Central Florida. Liverpool and Manchester are two historic cores and are formally two distinct metro areas, but are so interlinked they are arguably a single metro area, and are certainly a single multicore megaregion. There is contiguous suburban sprawl connecting the two cities with small gaps, and were British regional rail services better, their frequent urban rail networks would have touched. There are even some ties crossing the Pennines to Leeds; Britain has attempted to improve infrastructure between historic Lancashire and Yorkshire, using the language of megaregions to argue that this would boost the area’s economic profile.
Leapfrog urban connections
Western Germany and the Netherlands do not have contiguous sprawl in the same way that most developed countries do. On a satellite photo, the commuting zone of New York, Paris, Madrid, Toronto, or any other major city in their respective countries looks largely as a single blob of gray. The population density of this gray blob is higher in France than in the United States, but in both countries, a metropolitan area is made out of a single contiguous built-up area plus a handful of surrounding low-density exurbs.
In contrast, in Germany and the Netherlands there are undeveloped areas between adjacent cities. Most definitions of metropolitan agglomeration in Europe recognize that Cologne and Bonn are one metro area, but the two cities’ built-up areas barely touch and have farmland in between. The metro area of Frankfurt similarly contains multiple core cities with recognizable centers and some rural gaps between them, such as Darmstadt and Mainz. Urban areas with slightly bigger gaps do not necessarily fall into one metro area, but certainly comprise a single megaregion, including Germany’s largest, the Rhine-Ruhr with its roughly 11 million people and extensive internal S-Bahn connections.
Randstad is likewise a megaregion. The Netherlands zealously protects its high-yield farmland from urban sprawl, so suburbs are usually not contiguous with the cities they serve as bedroom communities for. There are agricultural gaps between Amsterdam, the cities of Flevoland, Utrecht, Rotterdam, and the Hague, and not too much commuting between the southern and northern edges of the combined region, and yet intermediate commuting and tight economic links mean it must be viewed as more than two or three disparate metro areas.
More controversially, I claim that the lower reaches of the Upper Rhine, from Frankfurt and Mainz up to Karlsruhe, form a single megaregion, and may even stretch farther up all the way into Basel. The gaps in urbanization between Frankfurt and Mannheim are not large – there is a city every few kilometers on both rail lines connecting the two cities. Moreover, the Frankfurt and Rhine-Neckar regions’ S-Bahns touch at Mainz, the Mainz-Mannheim line having recently been designated as S-Bahn quality and appearing on the regional schedules. The Rhine-Neckar S-Bahn in turn serves Karlsruhe. South of Karlsruhe the population density is high but less so, and the gaps between the cities are larger. But even without Baden south of Karlsruhe, the combined region has nearly 10 million people, and certainly has the highest GDP in Germany, as it is much richer than the Rhine-Ruhr.
Remember the Blue Banana?
In 1989, a group of French geographers led by Roger Brunet coined the term blue banana for a European megalopolis. As defined, it stretched from London or even Liverpool and Manchester in the north, across the Channel to the Low Countries, up the Rhine to Switzerland, and then across the Alps to Milan. The original definition deliberately omitted Paris from this zone, arguing that French urban geography was dominated by internal national links centered around the capital rather than the polycentrism of the Low Countries, western Germany, Switzerland, and Italy.
The last 30 years have not been kind to the Blue Banana. Much of Continental Europe was beset by a period of slow growth in the 1990s, sometimes called eurosclerosis; parts of it have slowly recovered in the 2000s and 2010s, most notably Germany, while others have stagnated, most notably Italy. In the 1990s, it was plausible to view Milan as more like Northern Europe than like Southern Italy. Today, it is no longer tenable. Before the 2008 crisis, Lombardy was as rich as Hamburg and southern Hesse and much richer than Stockholm and Copenhagen; today it is slightly behind Stockholm and slightly ahead of Copenhagen, and well behind Hamburg and southern Hesse.
The story of growth in the last generation has mostly been one of states, not regions. Northern Italy is much richer than Southern Italy, just as it has always been, but the entire country has equally stagnated. French growth has not been spectacular over this period, but it’s been better than Italian growth. Belgium, within the Blue Banana, has done better than France in the last generation, but not by much. In this entire period, the most notable subnational per capita income changes have been that London has pulled ahead while Northern England has stagnated, and that East Germany has grown faster than West Germany.
Megaregions and wealth
In the United States, the big megaregions have been loci of wealth, particularly the megalopolis. This has intensified in the current century. According to BEA data, since 2000, economic growth in the four core Northeast combined metro areas has exceeded the national average, gaining about 4 percentage points relative to the rest of the country in terms of both per capita income (from all sources) and net earnings (i.e. income from work). But even there, this is not the whole story, since Seattle, which is not in any megaregion, has had even faster growth.
Moreover, in Europe, there is no real correlation between megaregions and growth. The largest single megaregion in Europe, the Rhine-Ruhr, has slower economic growth than both the surging cities of southern Germany and the converging ones of the East. Paris and London are doing just fine as independent metro areas, Munich is still the richest city region in the EU, and Berlin is steadily converging to West German income levels.
Of course, no correlation and negative correlation are two different things. Just as the Rhine-Ruhr is slowly stagnating, the Frankfurt-Mannheim megaregion is growing, and Randstad has managed to recover from the recession alongside the rest of the Netherlands.
To the extent that there’s a link between megaregions and wealth, it’s that in developing countries, or even in midcentury America, poorer regions are mostly rural, and their cities tend to be small and less likely to interlink to form large metro areas. Thus, Eastern China has three megaregions with tens of millions of people each – Beijing-Tianjin, the Yangtze Delta, and the Pearl River Delta – underlying the wealth and urbanization of these regions; in contrast, the Indo-Gangetic Plain’s lower level of economic development means that even though population density from Bangladesh up the Ganges toward Delhi is as high as in southern Jiangsu, the cities are too small and too separated to form a Bangladeshi or West Bengali or Doabi megaregion.
But in a first-world context, the urbanization rate is about 100%. Even on-paper rural areas are within city regions and just happen to be small municipalities whose residents can drive in half an hour to a larger number of people than any premodern village pedestrian could interact with over a lifetime.
What this suggests is that the right way to think of first-world megaregions is not in terms of economic output, but in terms of density. In dense areas like the Netherlands, western Germany, England, and the Northeastern US, megaregions are likely to form out of links between adjacent cities. Not for nothing, the only part of the American Sunbelt where I’m comfortable describing metro areas as linking to form megaregions, Florida, also has the highest population density. The economies of Atlanta, Dallas, and Houston are a lot stronger than that of Central Florida, which is frankly a basket case, but cities in Texas and the Deep South are too far apart to function as megaregions.
Does high background density lead to higher incomes? Maybe. Strong urban networks really do allow for more economic specialization. But then these networks can be global, untethered from where one can travel by regional rail or urban highways. It’s an interesting question of economic geography, but on the level of a sanity check, some of the richest cities in Europe are doing just fine without the polycentric megaregional links going up and down the Rhine.
The American rail activist term regional rail refers to any mainline rail service short of intercity, which lumps two distinct service patterns. In some German cities, these patterns are called S-Bahn and RegionalBahn, with S-Bahn referring to urban rail running on mainline tracks and RegionalBahn to longer-range service in the 50-100 km range and sometimes even beyond. It’s useful to distinguish the two whenever a city wishes to invest in its regional rail network, because the key infrastructure for the two patterns is different.
As with many this-or-that posts of mine, the distinction is not always clear in practice. For one, in smaller cities, systems that are labeled S-Bahns often work more like RegionalBahn, for example in Hanover. Moreover, some systems have hybrid features, like the Zurich S-Bahn – and what I’ve advocated in American contexts is a hybrid as well. That said, it’s worth understanding the two different ends of this spectrum to figure out what the priority for rail service should be in each given city.
S-Bahn as urban rail
The key feature of the S-Bahn (or the Paris RER) is that it has a trunk that acts like a conventional urban rapid transit line. There are 6-14 stations on the trunks in the examples to keep in mind, often spaced toward the high end for rapid transit so as to provide express service through city center, and all trains make all stops, running every 3-5 minutes all day. Even if the individual branches run on a clockface schedule, people do not use the trunk as a scheduled railroad but rather show up and go continuously.
Moreover, the network layout is usually complementary with existing urban rail. The Munich S-Bahn was built simultaneously with the U-Bahn, and there is only one missed connection between them, The Berlin S-Bahn and U-Bahn were built separately as patchworks, but they too have one true missed connection and one possible miss that depends on which side of the station one considers the crossing point to be on. The RER has more missed connections with the Metro, especially on the RER B, but the RER A’s station choice was designed to maximize connections to the most important lines while maintaining the desired express stop spacing.
Urban rail lines rarely terminate at city center, and the same is true for S-Bahn lines. In cities whose rail stations are terminals, such as Paris, Munich, Frankfurt, and Stuttgart, there are dedicated tunnels for through-service; London is building such a tunnel in Crossrail, and built one for Thameslink, which has the characteristics of a hybrid. In Japan, too, the first priority for through-running is the most local S-Bahn-like lines – when there were only six tracks between Tokyo and Ueno, the Yamanote and Keihin-Tohoku Lines ran through, as did the Shinkansen, whereas the longer-range regional lines terminated at the two ends until the recent through-line opened.
The difference between an S-Bahn and a subway is merely that the subway is self-contained, whereas the S-Bahn connects to suburban branches. In Tokyo even this distinction is blurred, as most subway lines connect to commuter rail lines at their ends, often branching out.
RegionalBahn as intercity rail
Many regional lines descend from intercity lines that retooled to serve local traffic. Nearly every trunk line entering London from the north was built as a long-range intercity line, most commuter rail mainlines in New York are inner segments of lines that go to other cities or used to (even the LIRR was originally built to go to Boston, with a ferry connection), and so on.
In Germany, it’s quite common for such lines to maintain an intercity characteristic. The metropolitan layout of Germany is different from that of the English-speaking world or France. Single-core metro regions are rather small, except for Berlin. Instead, there are networks of independent metropolitan cores, of which the largest, the Rhine-Ruhr, forms an urban complex almost as large as the built-up areas of Paris and London. Even nominally single-core metro regions often have significant independent centers with long separate histories. I blogged about the Rhine-Neckar six months ago as one such example; Frankfurt is another, as the city is ringed by old cities including Darmstadt and Mainz.
But this is not a purely German situation. Caltrain connects what used to be two independent urban areas in San Francisco and San Jose, and many outer ends of Northeastern American commuter lines are sizable cities, such as New Haven, Trenton, Providence, and Worcester.
The intercity characteristic of such lines means that there is less need to make them into useful urban rail; going express within the city is more justifiable if people are traveling from 100 km away, and through-running is a lower priority. Frequency can be lower as well, since the impact of frequency is less if the in-vehicle travel time is longer; an hourly or half-hourly takt can work.
S-Bahn and RegionalBahn combinations
The S-Bahn and RegionalBahn concepts are distinct in history and service plan, but they do not have to be distinct in branding. In Paris, the distinction between Transilien and the RER is about whether there is through-running, and thus some lines that are RegionalBahn-like are branded as RER, for example the entire RER C. Moreover, with future extension plans, the RER brand will eventually take over increasingly long-distance regional service, for example going east to Meaux. Building additional tunnels to relieve the worst bottlenecks in the city’s transport network could open the door to connecting every Transilien line to the RER.
Zurich maintains separate brands for the S-Bahn and longer-distance regional trains, but as in Paris, the distinction is largely about whether trains terminate on the surface or run through either of the tunnels underneath Hauptbahnhof. Individual S-Bahn branches run every half hour, making extensive use of interlining to provide high frequency to urban stations like Oerlikon, and many of these branches go quite far out of the city. It’s not the same as the RER A and B or most of the Berlin S-Bahn, with their 10- and 15-minute branch frequencies and focus on the city and innermost suburbs.
But perhaps the best example of a regional rail network that really takes on lines of both types is that of Tokyo. In branding, the JR East network is considered a single Kanto-area commuter rail network, without distinctions between shorter- and longer-range lines. And yet, the rapid transit services running on the Yamanote, Keihin-Tohoku, and Chuo-Sobu Lines are not the same as the highly-branched network of faster, longer-range lines like Chuo Rapid, Yokosuka, Sobu Rapid, and so on.
The upshot is that cities do not need to neatly separate their commuter rail networks into two separate brands as Berlin does. The distinction is not one of branding for passengers, but one of planning: should a specific piece of infrastructure be S-Bahn or RegionalBahn?
Highest and best use for infrastructure
Ordinarily, the two sides of the spectrum – an S-Bahn stopping every kilometer within the city, and a RegionalBahn connecting Berlin with Magdeburg or New York with New Haven – are so different that there’s no real tradeoff between them, just as there is no tradeoff between building subways and light rail in a city and building intercity rail. However, they have one key characteristic leading to conflict: they run on mainline track. This means that transportation planners have to decide whether to use existing mainline tracks for S-Bahn or RegionalBahn service.
Using different language, I talked about this dilemma in Boston’s context in 2012. The situation of Boston is instructive even in other cities, even outside the United States, purely because its commuter rail service is so bad that it can almost be viewed as blank slate service on existing infrastructure. On each of the different lines in Boston, it’s worth asking what the highest and best use for the line is. This really boils down to two questions:
- Would the line fill a service need for intra-urban travel?
- Does the line connect to important outlying destinations for which high speed would be especially beneficial?
In Boston, the answer to question 1 is for the most part no. Thirty to forty years ago the answer would have been yes for a number of lines, but since then the state has built subway lines in the same rights-of-way, ignorant of the development of the S-Bahn concept across the Pond. The biggest exceptions are the Fairmount Line through Dorchester and the inner Fitchburg Line through suburbs of Cambridge toward Brandeis.
On the Fairmount Line the answer to question 2 is negative as well, as the line terminates within Boston, which helps explain why the state is trying to invest in making it a useful S-Bahn with more stops, just without electrification, high frequency, fare integration, or through-service north of Downtown Boston. But on the Fitchburg Line the answer to question 2 is positive, as there is quite a lot of demand from suburbs farther northwest and a decent anchor in Fitchburg itself.
The opposite situation to that of Fairmount is that of the Providence Line. Downtown Providence is the largest job center served by the MBTA outside Boston; the city ranks third in New England in number of jobs, behind Boston and Cambridge and ahead of Worcester and Hartford. Fast service between Providence and Boston is obligatory. However, Providence benefits from lying on the Northeast Corridor, which can provide such service if the regional trains are somewhat slower; this is the main justification for adding a handful of infill stops on the Providence Line.
In New York, the situation is the most complicated, befitting the city’s large size and constrained location. On most lines, the answers to both questions is yes: there is an urban rail service need, either because there is no subway service (as in New Jersey) or because there is subway service and it’s overcrowded (as on the 4/5 trains paralleling the Metro-North trunk and on the Queens Boulevard trains paralleling the LIRR trunk); but at the same time, there are key stations located quite far from the dense city, which can be either suburban centers 40 km out or, in the case of New Haven, an independent city more than 100 km out.
Normally, in a situation like New York’s, the solution should be to interline the local lines and keep the express lines at surface terminals; London is implementing this approach line by line with the Crossrail concept. Unfortunately, New York’s surface terminals are all outside Manhattan, with the exception of Grand Central. Penn Station has the infrastructure for through-running because already in the 1880s and 90s, the ferry transfers out of New Jersey and Brooklyn were onerous, so the Pennsylvania Railroad invested in building a Manhattan station fed by east-west tunnels.
I call for complete through-running in New York, sometimes with the exception of East Side Access, because of the island geography, which makes terminating at the equivalent of Gare du Nord or Gare de Lyon too inconvenient. In other cities, I might come to different conclusions – for example, I don’t think through-running intercity trains in Chicago is a priority. But in New York, this is the only way to guarantee good regional rail service; anything else would involve short- and long-range trains getting in each other’s way at Penn Station.
Many years ago, probably even before I started this blog, I visited family in Hamden, a suburb of New Haven. I took the bus from Union Station. When it was time to go back to New York, I timed myself to get to the bus that would make my train, but it rained really hard and there was no shelter. The time passed and as the bus didn’t come, I sought refuge from the rain under a ceiling overhang at a store just behind the bus stop, in full view of the road. A few minutes later, the bus went through the station at full speed, not even slowing down to see if anyone wanted to get on, and to get to my train I had to hitchhike, getting a ride from people who saw that I was a carless New Yorker.
Fast forward to 2018. My Brooklyn bus redesign plan with Eric Goldwyn calls for installing shelter everywhere, which I gather is a long-term plan for New York but one that the city outsourced to a private advertising firm, with little public oversight over how fast the process is to take. When I asked about the possibility of reducing costs by consolidating stops I was told there is no money for shelter, period. It was not a big priority for us in the plan so we didn’t have costs off-hand, but afterward I went to check and found just how cheap this is.
Streetsblog lists some costs in peripheral American cities, finding a range of $6,000-12,000 per stop for shelter. Here‘s an example from Florida for $10,000 including a bench. In Providence I asked and was told “$10,000-20,000.” In Southern California a recent installation cost $33,000 apiece. I can’t find European costs for new installation, but in London replacing an existing shelter with a new one is £5,700, or $8,000.
So let’s say the costs are even somewhat on the high American side, $15,000. What are the benefits?
I’ve found one paper on the subject, by Yingling Fan, Andrew Guthrie, and David Levinson, entitled Perception of Waiting Time at Transit Stops and Stations. The key graph is reproduced below:
The gender breakdown comes from the fact that in unsafe neighborhoods, women perceive waits as even longer than the usual penalty, whereas in safe ones there is no difference between women and men.
The upshot is that if the wait time is 10 minutes, then passengers at a stop with a bench and shelter perceive the wait as 15 minutes, and if there’s also real-time information then this shrinks to 11 minutes. If there are no amenities, then passengers perceive a 15-minute wait when they’ve waited just 6.5 minutes and an 11-minute wait when they’ve waited just 4. In other words, to estimate the impact of shelter we can look at the impact of reducing waits from 10 minutes to 6.5, and if there’s also real-time info then it’s like reducing waits to 4 minutes.
If the wait is 5 minutes then the impact is similar. With bench and shelter the perceived wait is 8.5 minutes, equivalent to a 3-minute wait without any amenities; with real-time information, the perceived wait is 6.5 minutes, equivalent to a 2-minute wait without amenities. There is some scale-dependence, but not too much, so we can model the impact of shelter as equivalent to that of increasing frequency from every 10 minutes to every 6.5 minutes (without real-time displays) or every 4 minutes (with real-time displays).
I have some lit review of ridership-frequency elasticity here. On frequent buses it is about 0.4, but this is based on the assumption that frequency is 7.5-12 minutes, not 4-6 minutes. At the low end this is perhaps just 0.3, the lowest found in the literature I’ve seen. To avoid too much extrapolation, let’s take the elasticity to be 0.3. Fan-Guthrie-Levinson suggests shelter alone is equivalent to a 50-66% increase in frequency, say 60%; thus, it should raise ridership by 15%. With real-time info, make this increase 30%.
What I think of as the upper limit to acceptable cost of capital construction for rail is $40,000-50,000 per weekday rider; this is based on what makes activists in Paris groan and not on first principles. But we can try to derive an equivalent figure for buses. On the one hand, we should not accept such high costs for bus projects, since buses have higher operating expenses than rail. But this is not relevant to shelter, since it doesn’t increase bus expenses (which are mostly driver labor) and can fund its ongoing maintenance from ads. On the other hand, a $40,000/rider rail project costs somewhat more per new rider – there’s usually some cannibalization from buses and other trains.
But taking $40,000/rider as a given, it follows that a bus stop should be provided with shelter if it has at least ($15,000/$40,000)/0.15 = 2.5 weekday boardings. If the shelter installation includes real-time info then the denominator grows to 0.3 and the result falls to 1.25 weekday boardings.
In New York, there are 13,000 bus stops, so on average there are around 180 boardings per stop. Even in Rhode Island, where apparently the standard is that a bus stop gets shelter at 50 boardings (and thus there is very little shelter because apparently it’s more important to brand a downtown trunk as a frequent bus), there are 45,000 weekday riders and 3,000 stops, so at 15 riders per stop it should be fine too put up shelter everywhere.
The only type of stop where I can see an exception to this rule is alighting-only stops. If a route is only used in a peak direction, for example toward city center or away from city center, then the outbound stops may be consistently less used to the point of not justifying shelter. But even that notion is suspicious, as American cities with low transit usage tend to have weak centers and a lot of job and retail sprawl. It’s likely that a large majority of bus stops in Rhode Island and all stops within Providence proper pass the 2.5 boardings rule, and it’s almost guaranteed that all pass the 1.25 boardings rule. And that’s even before consolidating stops, which should be done to improve bus speed either way.
At least based on the estimates I’ve found, installing bus shelter everywhere is a low-hanging fruit in cities where this is not already done. In the situation of New York, this is equivalent to spending around $550 per new weekday rider on transit – maybe somewhat more if the busier stops already have shelter, but not too much more (and actually less if there’s stop consolidation, which there should be). Even in that of Providence, the spending is equivalent to about $6,600 per rider without stop consolidation, or maybe $3,000 with, which is much better than anything the state will be able to come up with through the usual channels of capital expansion.
If it’s not done, the only reason for it is that transit agencies just don’t care. They think of buses as a mode of transportation of last resort, with a punishing user experience. Cities, states, and transit agencies can to a large extent decide what they have money for, and letting people sit and not get drenched is just not a high priority, hence the “we don’t have money” excuse. The bosses don’t use the buses they’re managing and think of shelter as a luxury they can’t afford, never mind what published transportation research on this question says.