The Interurban Option – Prereqs

Alon‘s had a long-running series on the comparative construction costs of transit modes in different countries. It’s good reading.

At the same time, I’ve always been a fan of the “Interurban Option” – running commuter-rail levels of service to low-density areas using rolling stock that is compatible with inner-city LRT systems. I’ve previously suggested that North Carolina’s research triangle would make an ideal location for such. But in order for lower-ridership/lower-frequency LRT to work, you need lower-cost LRT.

Chicago has a long history of interurbans operating over the “L” network. The North Shore came in on the Howard line, while the Roarin’ Elgin got on the Congress “L” at Forest Park. These trains operated all the way to the center of the Loop, as contrasted with other cities where interurbans ended at suburban terminals (e.g. Philly-69th) or duplicated the streetcar system (e.g. Pacific Electric). This kinda thing should work.

Back in the 1960’s, the CTA retrofitted some former interurban trackage into an “L” extension – the Skokie Swift, or yellow line. Construction was ridiculously cheap. Catenary had been pulled down when the North Shore Line folded, but a few miles of third rail remained extant. Rather than rebuild catenary over the entire route, wire was hung only over the unelectrified portion. To switch, trains built up speed, coasted past the end of the third rail, and raised a special bespoke pantograph designed by the PM, to switch to wire. This continued until 2004.

This sort of inexpensive construction wasn’t unheard of in the 60’s and 70’s. The San Diego Trolley originally used street-level boarding (this was pre-ADA). Calgary’s C-trains didn’t have air conditioning until quite recently (and some still don’t). Portions of Portland’s MAX and Baltimore’s LRT were single-tracked, though the latter proved short-sighted.

Similar cost-containment would be needed to make Interurbans a success. For instance, the Skokie Swift carries 7,000 riders daily. At current prevailing costs, a double-track LRT line of the Skokie Swift’s length with an ultimate ridership of 7k would have to be considered a failure, not worth the effort, fodder for a Randal O’Toole piece on the superiorities of bus rapid transit. But 7,000 riders would be quite respectable for a single commuter rail line. Many of the legacy Philly lines operate in this region, while New Mexico’s Rail Runner carries only 4500.

So let’s talk about what it would take to make Interurban – LRT vehicles at commuter rail frequencies – work.


This one’s pretty simple. Wooden ties don’t last as long as concrete, and the heavy federal match on most rail projects biases everything in favor of high capital cost/low operating cost. But wooden ties are *cheap*. This is the reason historically pretty much every railroad everywhere has used them. Moreover, a rail tie which sees 15-to-20-minute peak headways and 60-to-90-minute headways elsewhere is going to be subjected to a lot less stress than a core LRT trunk running every 7 minutes. And while the “best practices” recommendation in your project plan (and associated cost benefit analysis) is going to assume replacement on 20-year intervals, deferred maintenance as practiced by all legacy US operators will see lightly used wooden ties pushed out to a 40 or 50 year lifespan.


Most US LRT goes in with catenary befitting a German mainline railway. This is, again, an outgrowth of the US transportation culture’s bias towards front-loading capital costs. Trolley wire is cheaper, but wire is more difficult to keep properly tensioned. The increased friction from slightly slack wire increases wear on both wire and pantograph. However, for a system that runs 90-minute off-peak headways, the  additional maintenance costs are trivial compared to the cost of the primo stuff.


You can start to see the common dynamic here. Wood is cheaper than steel, but doesn’t last as long. What’s better? Electric utilities public and private invariably choose wooden poles. Only with trains do we start with stainless.


Trains are enclosed spaces. LRVs are more enclosed than commuter trains or Amtrak, since they’re not vestibuled. Being on an LRV stopped in the middle of nowhere can be somewhat disconcerting. Moving all meets to stations means the train waits for a couple minutes with the doors open (or if it’s hot outside, with the doors openable by pushing a button). This is much more pleasant for the passenger.

Moreover, delays can and do happen. Skimping on passing tracks based on an operating plan will only lead the system to be obsolesced sooner. A stop every couple miles with a passing track is good for 10-minute headways. A stop every mile with one is good for a bit over 5. This sounds like overkill until the 5:53 gets delayed by a fair trade protest until the 6:07 and the 6:32 are right behind it.


Pedestrian overpasses on low-frequency rail lines can be justified where there are multiple tracks, which are also used by 79mph freight trains that take eight miles to stop. However, most every LRT out there is capable of a 3.0mphps deceleration rate or better, which lessens the need for ped grade separation.

With at-grade ped crossings, island platforms are much safer because there is no “multiple threat” issue. The platform itself serves as a pedestrian refuge, and if there is access from both sides a “zig zag” sidewalk can be used so that all peds cross in front of or behind the train. Different agencies appear to have different preferences in this regard. Island platforms are also optimal from an operations perspective, since you can “wrong rail” trains without any inconvenience to passengers. Among other things, this allows express trains to pass locals.


In practical terms, this means that a single track should be built off-center, so that it becomes part of a later double track. At a minimum the distance off the ROW centerline should be half the minimum track spacing (e.g. 6 to 7 feet), but ideally there’d be enough room for a second track under construction as well as a row of construction vehicles. This probably means grabbing a 10′ easement along the property adjacent to the future second track. In most suburban contexts, the cost of this easement will typically be a fraction of a second track, even if you factor in the the discounted cost of a second track 20-30 years in the future.


In theory, running next to rail corridors is a great idea. In practice, this only works if the adjacent land uses don’t have rail service. Otherwise you end up like Dallas or San Jose, building gigantic multi-mile concrete bathtubs to take LRVs over industrial leads that see one train every three weeks.


A lot of suburban commercial strips have relatively continuous boundaries where the backs of the stripmalls and big boxes end and single-family homes behind. Teasing a rail line through this boundary region requires a ton of strip takes and probably some people’s backyards, but once this is done you’ve got grade crossings every 1/4 to 1/2 mile and there’s enough queue storage space between the gates and adjacent traffic signals that you don’t need to spring for a fiber interconnect. The rest of your design takes care of itself.

Similar routes exist along old land survey lines. Master-planned communities of the Woodlands/Highlands Ranch/Summerlin variety tend to make up only a fraction of the urban form. The rest of it is built out piecemeal, so that the original outlines of 50-100 acre plots can be seen from the air. In the Mountain West and the Plains these are often half- and quarter-sections; in Texas and the East they’re a little more esoteric. Following these lines you can grab parking lots, yards, etc. Garden apartment complexes are particularly easy to work with since you can take out the strip of parking along the fence, raze a building, then reconstruct the parking in its place.

All of this drives up right-of-way costs, but you also get to minimize capital and O&M outlays. And the cash saved by using a “free” roadway or highway ROW is easily wiped out by the cost of median paving or overpasses at interchanges. Find an alignment that lets you use cheap construction, instead of trying to find a cheap alignment and then letting the construction cost chips fall where they may.

Will the US ever see non-traditional transit?

I was discussing transit with an acquaintance today, and mentioned monorail as one technology which could be used on certain corridors. This person stated that they thought monorail was “kitschy” and, perhaps realizing that subjective perception of aesthetic merit is a poor criterion for mode selection, proceeded to reach for other reasons to justify writing off monorail as a technology entirely.

I find these conversations to be somewhat aggravating, as the objections are the same ones I heard as an intern with the Seattle Monorail Project a dozen years ago.

“Monorails aren’t a proven technology”

Japan has had monorails in operation for nearly 50 years.

“Monorails can’t carry high numbers of people”

Tokyo-Haneda carries 120,000 a day. Chongqing carries 1.1 million on a system which is roughly a 50-50 split between monorail and subway.

“It takes a long time to switch trains”

No, segmented switches in use in Chongqing and Tokyo cycle in 5-6 seconds. And the suspended Shonan Monorail operates on a single track, relying on intermediate switches to maintain 8-minute headways throughout the day.

I used to field these questions all the time at public meetings. At the time, we thought their days would be numbered. Seattle would have its system, there would be a heavy urban transit monorail in operation with the US, ignorance would no longer be possible. Alas, the Seattle political process struck in the way that it always has, and after four votes in favor of the monorail a fifth took it away. (This isn’t new – in the 60’s, the Feds were prepared to cover most of the cost of a heavy rail system for Seattle. It was rejected, and the money was sent to Atlanta, where it created MARTA).

Meanwhile, Vegas built their line. And while it has decent ridership density (12,000 people over a four mile route), its location among the casinos does little to establish it as a “serious” transit line in the minds of people for whom “seriousness” matters as a criteria independent of such things as system capability, capital or O&M cost.

I was thinking about this when it occurred to me: this sort of myopia isn’t unique to monorail. In fact, it has afflicted many transit modes.

Consider, for instance, Skybus. Skybus was first-generation AGT, developed from scratch by Westinghouse. Automated, driverless, rubber-tired trains would run every 6 minutes, replacing Pittsburgh’s decrepit streetcar system. Skybus came very close to being a reality, but was ultimately killed by political infighting. Pittsburgh instead got a couple of light rail lines and a busway that no one rides, while the rest of the streetcar system was torn up.

But that wasn’t the end of the technology. While Westinghouse was unable to sell it to the transit operators, they did have some success with airport operators, installing the technology at SeaTac (1969) and Tampa International (1971). Over the years the technology passed through Adtranz to Bombardier, where it continues to be the predominant AGT design in the airport market.

Alas, Bombardier never re-marketed rapid-transit-sized rubber-tired AGT, likely because they also have the rights to the original Skytrain technology. But latecomers in other countries were successful; France’s MATRA was able to sell its VAL system to Lille and Toulouse, in addition to various airport clients. The technology passed to Siemens, who have installed line-haul AGTs in Taipei, Seoul, and Turin. Rubber-tired AGTs also made it big in Japan, with the Kobe Port Liner, the Saitama New Shuttle, the Nippori-Toneri Liner, Yurikamome Line and Yokohama Seaside line all carrying substantial passenger loads.

Up in Canada, Vancouver has continued to build out their Skytrain system. But when the contract for the most recent line was tendered, rules were adopted that expressly prohibited consideration of cost savings resulting from using the same technology as the rest of the system. Thus in 2009 the Canada Line opened not with steel-wheeled LIM-powered AGT, but with regular-ass third-rail powered EMUs. Not really all that different from the tech of 100 years ago. Meanwhile, the only substantial implementations of the Skytrain tech outside Vancouver have been in Beijing, Seoul, and Kuala Lampur.

You see the pattern here. Southeast Asian cities look at a rapid transit line, consider all the technologies available, and then make a decision. But Americans (and to a lesser extent Canadians) seem to first run transit decisions through a filter which excludes anything that’s not “serious,” meaning anything that isn’t substantially identical to the transit technology that was installed in cold, dreary east coast cities a century ago.

So Seattle adopts LRT and then ends up with a giant concrete bathtub where they could’ve had two slender monorail beams instead. Pittsburgh adopts LRT and then never expands it, because the city is too hilly to do so without new tunnels. And all across the US, we’re having very earnest discussions about streetcars, which have an average speed of approximately two miles per hour and can generally be outpaced by everyone from novice cyclists to Rascal scooter owners. Meanwhile, Kuala Lampur’s monorail is so popular that they’re extending the platforms to allow longer trains.

Why is this so?

The Spur needs to split at Shepherd.

One of the most reliable backups inside the Loop right now is the PM reverse-commute inbound on 59. In the early afternoon, when most of the Transtar map is green, the Shepherd-Spur segment is yellow. By the height of the rush, the queue extends at least to Buffalo Speedway. This happens for a number of reasons – but first, the fix. Because it’s ridiculously simple.

Move the diverge point to Shepherd.

To see why, here’s 59 in Greenway. Five lanes. Now here’s 59 entering Midtown, a couple miles further to the east. Again, five lanes – and about to lose two of them.

So add a couple miles of barrier and put the split at Shepherd. If you’ve driven this route once, you can see how it would immediately shave several minutes off the trip to Downtown/Midtown/Montrose. The Spur is never jammed in the reverse direction, so anything that effectively lengthens the Spur lengthens the distance of hassle-free 60mph cruising. But such a configuration would also help drivers continuing on 59. Here’s why.

Why does 59 back up?

In short, three tailbacks and a weave. 59 through Midtown consists of three lanes which split into three freeways, and backups on any of these are telegraphed through onto the mainlines. The splits:

(i) The two left lanes head to 45. While there is a predictable queue to get onto the Pierce Elevated, it rarely backs up onto 59.

(ii) The two right lanes head to the Eastex Freeway, continuing as 59. This is a more common location for congestion, as there are two lanes coming off 59 and two lanes coming off 288, which narrow to three past the GRB.

(iii) An auxiliary lane gets picked up from San Jacinto which peels off onto 288 after a few hundred feet. 288 South is often backed up, and any tailback from 288 will cause the weave from San Jac to break down, effectively taking out two lanes. This by itself is sufficient to back up 59 past Shepherd, even if both the Eastex and the ramps to 45 are relatively clear.

Even when all three destination freeways are running smoothly, the short weave from San Jac still causes a reduction in capacity that is available through Midtown.

How does moving the Spur diverge point westward help? Because the Spur merge acts as a repeater for the Midtown bottleneck. In a normal traffic bottleneck you have three zones. You have the bottleneck itself, you have a queueing section where traffic stacks up waiting to get through the bottleneck, and then you have a free flow section upstream of there.

As traffic backs up on 59 past the Spur, drivers are faced with a dilemma. Do I sit and queue here in the right three lanes, which aren’t moving? Or do I get over and zoom past until right before the split? Many, understandably, choose the latter. But what this does is create a new bottleneck at the point where the Spur diverges, because traffic is merging into the left lane and then trying to cross over to get to 288 or stay on 59.

Wouldn’t a Shepherd Spur Split just move the repeater point to Greenway? No, for two reasons. First, a lot of the time the backup from 288/59/45 will never make it to Shepherd. Cars will happily (or not) queue up in the right three barrier-separated lanes, and the end of the queue will be somewhere between Midtown and Shepherd. And when it does make it back there, it still won’t be as bad, because there are already auxiliary lanes between exits in Greenway.

The Limits of Cute Transportation

This is a draft I saved back in November and then forgot about. I figure I must’ve accidentally published it for a few minutes, because more than one person has emailed me asking where it went. So, here it is.

So here’s the east approaches to the Marquam Bridge, in Portland.

Roll up to a long-range planning meeting and you’re liable to hear idle talk about tunneling or removal or somesuch. It doesn’t matter what the actual likelihood of this happening is. Dislike of the Marquam bridge is sort of a litmus test for serious Portland movers-and-shakers. Not only is it a freeway, it’s ugly. Or that’s the official line anyway. What’s not ugly? The Streetcar.

Cute. With a cutesy sign in the background to emphasize the cuteness. But, not particularly useful. The streetcar is actually slower than equivalent bus routes (though the existence of rail bias produces a decent amount of ridership) and is best thought of as a land development tool. What does a useful, modern rail line look like?

Super useful. Carries nearly 800,000 people a day. Trains reach 60mph between stops. The Metro is the very key to the District’s success; without it, you simply couldn’t have continued to locate major government and office buildings inside the Beltway, and the residential blocks would not have followed either. But note that we’re way outside of “cute” now, and into dour 70’s modernism. Exceedingly well-executed modernism, which has held up surprisingly well. But not cute.

What does an even higher level of service look like?

That’s the Sanyo Shinkansen, a couple miles outside of Shin-Osaka. That’s about a 1km radius curve, acceptable only in close proximity to station where all trains stop. Get a bit further out and the curves are substantially broader. But now we’re back to the Marquam bridge, we’re back to the “ugly.” The sort of thing that all right-thinking NIMBYs would oppose as a blight on the landscape.

When you read criticism of urban highway networks, you see a fair bit of criticism that highways and cars are “out of scale” and “don’t fit” in an urban context. This kind of talk is synonymous for “highways are not cute.” Well, quality rail infrastructure isn’t cute either. If you’re triggered by the sight of a multi-deck freeway interchange, you’re not prepared for high speed rail either. Cuteness shackles you to forever move at the speed of surface light rail.

So you need to have the ability to rise above cute. Houston does. Vancouver and Toronto do, but Montreal doesn’t appear to. New York used to, but whether they can still go big and brutalist remains to be seen.

To keep improving the transport network, you need a citizenry that’s cool with concrete. And it’s this reason why I’d wager we’ll see at least three true high-speed rail systems in operation (California, Texas, and a third – maybe Florida, maybe NC, maybe DesertXpress, maybe electrified Cascades) before we ever see major capacity expansions in the Northeast, such as an inland route from NYC to Boston via Hartford. Beyond the capital and political issues, you have a citizenry that isn’t equipped to appreciate the aesthetics.

Riding the Texas Central

A commenter asked for my opinion on the Texas Central Railway proposal for high-speed rail, and the various challenges facing it. Thought it merited its own post. So, my thoughts.

My first feeling is skepticism. High speed rail pairs extremely low operating costs with extremely high up-front capital costs. Debt service thus becomes a huge determinant in whether or not a line can “break even.” A few basis points can make or break the entire system.

Privately-financed HSR infrastructure ends up costing a lot more because private borrowing costs are so much higher. True, a private system may see lower construction costs, by being exempted from federal contracting and prevailing wage requirements. But in a right-to-work state like Texas, the prevailing wage isn’t all that steep to begin with. And even states with a history of iffy finances (e.g., California) can issue bonds cheaper than most companies.

Even a system with good ridership can thus operate “in the red” if the banks are extracting too large of a cut. This happened in Taiwan. Three years after their HSR system opened, they were bleeding money in interest payments, and had to be refinanced with a new set of loans backstopped by the Bank of Taiwan.

There is a workaround – namely, the RRIF. RRIF lets private railroad entities get loans for whatever the going rate is on T-bills, thus avoiding a Taiwan scenario. As you might imagine, the Republican and Libertarian establishments hate this program. For instance, here’s Reason coming out against the Los Angeles-Las Vegas line, and here’s Heritage echoing the sentiment.

So my thoughts are, given that the exact same people who would oppose a publicly-financed and owned rail system are going to be just as opposed to a privately-financed and owned rail system with government loans, why not just go for the full enchilada? Build it with public funds and then let JR or First or Veolia contract out to run the trains.

Having expressed skepticism of the financing model, my second reaction is excitement. I’m a partisan. I want to see 700-series Shinkansen in Texas. The reason is loading gauge. Off-the-shelf Euro trainsets are built to Berne Gauge, which is a bit over 10′. Existing North American trainsets are built to Plate C, which is a bit wider. But Shinkansen uses its own loading gauge, explicitly designed for HSR.

So while the ICE-3 is 9’8″, Amfleet is 10’6″, and the “international” Velaro D is 10’8″, the N700 clocks in at 11’1″. Do you shop at Casual Male? Do you drive an SUV? This is the train for you. Of course, the Japan/Taiwan spec has 5-across seating in economy class. But I have to imagine that no US operator would be stupid enough to bring that here. And indeed, Texas Central’s website shows four-across “green car” seating. An all-Green Car Shinkansen would really be quite something.

Moreover, when it comes to the route structure, my attitude is one of endorsement. Texas Central proposes to put one station in downtown Dallas and one station on the northern outskirts of Houston. This makes sense when you consider demographics and airport location.

On the demographic side, HSR ridership skews toward a higher-income, service-sector crowd. Dallas’s office sprawl heads northward, which tends to funnel the Plano-Frisco-Galleria folk through downtown. By contrast, Houston’s office sprawl goes west, and Dallas-bound traffic utilizes any of several ring roads before finally coalescing somewhere in The Woodlands. This supports a Downtown Dallas station and a North Houston station.

On the airport side, Houstonians have to drive a ways out to reach the aeropuerto, regardless of what flight they’re taking. But Dallas folks have Love Field, which is just a couple miles from Downtown. An outer Houston rail station is thus competitive with air travel in a way that an outer Dallas one wouldn’t be.

As for the issue of stations in Fort Worth and DFW, my tone is one of exasperation. Direct DFW service is a great idea, but there is simply no cheap way to do it. You’re either tunneling under the runways (mega expensive) or extending the APM to Centreport. And since Skylink operates within the secure area, you’d need to construct a satellite terminal for check-in and baggage handling.

Personally, I’d go for a tunnel. It works at Narita, it works at CDG. But that’s some major capital outlay and it seems crazy to me to hang the success or failure of a Houston-Dallas train line on service to the airport. It also doesn’t escape my attention that, assuming the system is ultimately expanded to include the rest of the triangle and beyond, a direct DFW connection tends to advantage airlines with a DFW hub (e.g., AA) at the expense of those using IAH or Love (United, Southwest).

In summary, then, my short take on it is this: I hope they pull it off.

Haussmannization Continues, or, Protesting TOD

This weekend, a mile and a half of the Odakyu system went underground.

The old line had level crossings every few hundred feet, as seen in Google. The undergrounding is part of a long-range plan by Tokyo’s Setagaya Ward to redevelop the area around Shimokitazawa Station, a major interchange point between two railway lines that connect to two of Tokyo’s commercial districts. Think Secaucus Junction, but with a low-rise neighborhood in place of the Meadowlands.

Phase II of the plan is to construct a Transit Center on the right-of-way vacated by the Odakyu Line. This gets translated to English as “rotary plaza,” but the concept is the same – a big concrete apron with pullouts, shelters, and enough space for a full-sized bus to turn around. The transit center will be accessed by a new street cut through the neighborhood’s Tokyo-standard warren of alleyways. It’ll have 2-3 lanes (one each way, plus turn bays), wide sidewalks, and landscaping. In cross section it is almost identical to Preston Street near Dean’s Credit Clothing.

Near the rail station, and along the new street, building height limits will be increased to 200 feet. Put together, the plan looks like so:

Not exactly a bad redevelopment plan, eh? Take a major railway station, make it a focal point for bus service, allow higher-density development. Great.

But you wouldn’t know this from reading the English-language media. For instance, a Global Voices article states: “the plan will split Shimokitazawa apart with a 26-meter wide expressway.” This brings to mind an elevated four-lane highway, not a two-lane surface street with a 26m total right-of-way. A New York Times piece also sticks to the divided highways theme, stating: “A shadow has fallen straight across the heart of this pulsing neighborhood. In four years, city officials plan to start building an 81-foot-wide thoroughfare that will slice Shimokitazawa in two.”

So why the misrepresentation? A closer read of the NY Times article suggests an answer. “I couldn’t believe it,” said Kenzo Kaneko, 41, an architect who lives here. “They just announced the death of the neighborhood, without asking us what we thought.”

Longtime readers will recall I’ve previously outlined a four-step process by which neighborhoods are gentrified.

1.) Poor creatives (e.g., artists)

2.) Affluent creatives (e.g., architects and marketers)

3.) Affluent people who like to think of themselves as creative (e.g., MBAs)

4.) Civil Engineers

Shimokitazawa, it appears, is on the tipping point between phase II and phase III. 17-story apartment blocks, with all of the latest amenities, will further open up the neighborhood to muggles. They must therefore be stopped.

Were this article being written about San Francisco or Berlin, I might start talking about zoning. But Japan largely does not practice Euclidean zoning. What Tokyo and other cities do have is density, FAR, and height limits. That’s what comes into play here.

It’s therefore useful to note that the same basic dynamic – neighborhood sees restricted development, people become used to restricted development, huge outcry results when restrictions are proposed to be lifted – occurs in non-Euclidean systems as well. Whenever architectural students from Rice or wherever get together to brainstorm what sort of exciting “form based codes” or other land use restrictions they might lay down on Houston, they should keep in mind the story of Shimokitazawa. Any restrictions on land development at all will inevitably lead to a citizenry that feels entitled to preserve their neighborhood in amber. Densification will be that much harder, with associated increases in sprawl and reductions in housing affordability.

Just don’t do it.

As an aside, here’s the page of the official protest group, where they list their alternative proposals for redevelopment.

Near, Far, Wherever you are?

Open question to Houston readers. You’re on the road. You need to make a left or U-turn. Do you pull to the near side of the median opening? (blue car) Or the far side? (red car)

On the one hand, the near side movement is common to almost all our signalized intersections. On the other hand, the far side movement is common to rural divided highways, urban divided highways in many other states, and Houston arterials of yore. H-town thoroughfares were divided going back to the 50’s, but the bullet-nose median opening – which enables the nearside movement – didn’t really become omnipresent until the 80’s.

What do you do?

A Super Cheap Muni Capacity Expansion

One of my Portland friends recently posted an article entitled 25 things I wish I knew before moving to San Francisco. Astronomical rent is #3, illustrated by a map of rents by neighborhood. The blogger correctly fingers housing restrictions, linking to a Pando article which in turn links to this SF Public Press piece which is basically a bunch of people concern-trolling about whether SF can add 30% more people in 30 more years – an annualized growth rate of less than 1% per year.

I’d called this the “captain, I’m givin’ it all she’s got!” argument against future growth. And frankly, it gets old. You can build new sewer lines. Portland did. But it got me thinking.

Previously on this blog, I’ve argued for concentrating upzoning in the Sunset district. But if SF were to do this, there’d need to be a massive improvement in transit capacity through Twin Peaks. Here’s a cheap option.

Upgrade Twin Peaks Tunnel to allow 2-minute headways

Set up the signaling to allow solid 2-minute headways between West Portal and Embarcadero. (Realistically, this means a capacity for 90-second headways, to allow for recovery from delays).

Run six-car trains

If Seattle can do it, so can SF. LA runs six-car trains everywhere, and LA’s rolling stock is 22% longer than SF’s.

Switch to high platforms everywhere

This is the “low hanging fruit” as far as dwell times are concerned. When combined with 6-car trains, this would require converting a fair number of cross intersections to right-in, right-out. But platforms and intersection realignments are nothing compared to, say, a new subway.

Cannibalize Market

Currently all of the Muni Metro lines go into the Market Street subway, and then heritage trolleys operate at the surface. The heritage trolleys could be kicked off Market (relegated to the F line), freeing the surface tracks for additional LRT capacity. Routing the J-Church to the surface would free up some space in the subway. But that might not be enough. So…

Consider ending the J-Church in the Castro

There’s an existing “Y” at 17th and Church and there’s a U-Turn at Castro and Market. This would allow the “J” to be recast as a shuttle. Which gets you a very nice…

Operating Pattern

All N-Judah trains run via the surface tracks on Market. All J-Church trains terminate at Castro. L-Taraval trains alternate with either M or K trains in the subway and Twin Peaks Tunnel. This yields:

3-minute service on the N-Judah

4-minute service on the L-Taraval

8-minute service K-Ingleside and M-Ocean View

Compared to current service, this is:

A 135% frequency increase for the N-Judah (8-9tph to 20tph)

A 90% frequency increase for the L-Taraval (8tph to 15tph)

No frequency increase for the K-Ingleside and M-Ocean View (7-8tph to 7.5tph)

Present service operates with 2-car trains on the N and L and single-car trains on the K and M, which are coupled/decoupled at West Portal. When frequency increases are combined with 3-car trains throughout the system, this yields:

A 250% capacity increase for the N-Judah

A 180% capacity increase for the L-Taraval

A 300% capacity increase for the K-Ingleside and M-Ocean View

And that’s with some platforms, some partial street closings, and some signaling upgrades. I think SF City could accommodate another half-million rather easily.

A simple model for vertical mixed-use vacancy.

Matt Yglesias is all “gee, why do storefronts below new condos go unused?

Well, let’s see. There’s two things to understand here. The first is that at many businesses are not viable even at zero rent. Let’s say my uncle dies and wills me his panel van. I could sell donuts out of this van! Have my own business! So I read up and I see that Houston and Dallas are both large markets for donuts and I decide to park my van in a field halfway in between. Is this a viable business strategy?

No, it’s not. For one thing, it’s very likely I’m going to lose money on the donuts. Presuming that I purchase wholesale donuts from the baker at $X and sell retail donuts to the public at $Y, then every day I need to sell at least Z donuts, where Z is Y * (number of donuts purchased / X). Then there is the opportunity cost of spending all day in a relatively unprofitable donut truck. I just as easily could be delivering cupcakes for $10 an hour.

The second thing to understand is that Urban Planners really, really, really like vertical mixed-use. It provides clear visual rebellion against the earlier generation of planners who really, really, really liked single uses, preferably as well-separated as possible. In many places planners will give you a “bonus” for providing retail spaces under residential. Maybe extra height, extra FAR. In other places planners require that you add mixed-use. Dallas did this, and as a result they have some very pretty streets that provide actual, built examples of the sketches you see in planning documents about the “transformation” that will happen to a certain street after bike lanes are added.

Sometimes it works.

The problem is, if you provide a bunch of financial incentives for the construction of retail space, you’re going to get a lot of retail space. And if you get enough of that, eventually you’re going to drive prices so low that it’s only rentable to unsavory businesses. There is a reason why “low rent” and “unsavory” are often synonyms. Say you’re looking to open a new Banana Republic store. Certainly, you’d like to pay next-to-nothing for a lease. But the places where you can find a next-to-nothing lease aren’t generally the places where it would be profitable to operate a Banana Republic. Conversely, given the obscene markups in porn it would be quite possible for a store in the Galleria to be profitable, and indeed there’s one next door. But in general there’s no benefit to locating a porn store in a major mall, you’d just as soon pick some place cheap like Sharpstown or Tidwell and 45.

Now if you own a stripmall, this is not a problem. You lease to whoever’ll pay. Five years ago in the Montrose you had a leather shop on Westheimer at Stanford and then you had a porn store three blocks east at Mason. With rising rents those tenants left and have been replaced with yet another tattoo parlor and a booshy wine bar. In theory, the glut of commercial space in San Diego could follow the same trajectory. Except SD’s glut of commercial space isn’t in stripmalls, it’s in residential blocks. Any commercial use poses a nonzero externality to people who live on top of it. And if the Slate comments section is to be believed, the HOAs for the condo buildings largely control those spaces.

What you get is a curve that looks like this.

As the “market price” for the rental space declines, so too does the opportunity cost. A retail space which could go for a couple grand a month not leased is a couple grand a month out of your pocket; one that only goes for 500 is only 500 lost. Conversely, as demand for the space goes down, the externalities of potential tenants increase. The clothing store doesn’t impose many externalities, but bars do, and there’s only room for so many clothing stores in a given yuppie neighborhood. The intersection of the two curves is the point at which it’s “not worth it” to lease the space. For a condo board this would basically be the point at which a given commercial use causes sufficient distress to the residents to negate the rent it brings in.

How do you move these curves? Well, in this relatively simplistic model, you don’t. Rather, what you do is stop requiring developers to build vertical mixed-use. If I walk downstairs in my “single use” apartment block and round the corner to a “single use” commercial storefront, I have not gone any farther than if my apartment was on top of that storefront. But placing the uses on separate parcels simplifies building codes, insurance, ownership, and leasing.

Removing the financial incentives for ground-floor retail can raise “market rents” (the theoretical rent at which all the spaces could be filled) by causing less retail space to be built in the first place. And as market rents increase the quality of the tenants will naturally also increase so that the externality cost goes down. At some point supply will be reduced to the point that rents exceed externalities at which point all the space is leased.

In a city like Houston you have no financial incentive for mixed-use and in fact the relatively inflexible parking requirements tend to militate against mixed-use, and even then you have Post Midtown which is some very nicely developed VMU. But it is important to note that only a part of the Post Midtown street frontage is actually VMU; much if it is quite residential. And Post Midtown, in turn, is only a small portion of the total neighborhood housing stock. Camden has developed more units and is institutionally skeptical of vertical mixed-use. That can still yield some nice buildings, but it’s not the sort of “activated streetscape” that’s hot in urban planning right now.

So this is, at root, a failure of regulation. Except rather than planning and zoning regs that favor low-density suburban districts, it’s a reg that in theory favors walkable urban districts. And who knows, perhaps foot traffic in that part of San Diego will eventually rise to the point where those spaces become leasable to the type of tenants the residents want. However, in order for that to happen, more people will need to move to the neighborhood, and newer buildings will need to not include retail space. As long as more condos = more retail, the balance will never change, and that neighborhood will have a permanent glut of retail. Cities with this problem have no choice but to stop incentivizing it. Which is to say, they need to do it the way Houston does.

Establishment Libertarianism finds the frequent bus.

Reason Magazine, which is gung-ho on food trucks but hostile towards the mass transit and zoning policy that creates the sort of neighborhoods in which food trucks thrive, recently moved into a new headquarters. The building is in a little finger of City of Los Angeles proper that pokes the enclave of Culver City.

It is a ridiculously transit-accessible site. The building sits across from the turnback for LA Metro route 108, which runs 15-minute headways or better all day. It’s a busy enough route to justify an express overlay, the 358, although the peak-direction-only service isn’t particularly useful to Reason workers. It’s also a short, 3-minute walk from the Route 110, as well as several Culver City Bus routes, which are somewhat less frequent. Both Metro routes provide easy connections to the Blue Line light rail.

Availability doesn’t necessarily mean usage, however. You have to wonder if bus riding at Reason might fall into that nebulous category of “things you can do, but probably shouldn’t, since it might be looked down upon by your superiors.” But it is still interesting regardless. The establishment L’s seem to have generally done a better job than the R’s at attracting younger folks into the fold, as well as convincing their elders to dress in a manner approximating what they think younger folks dress like (witness the continuing adventures of The Jacket, the sentient lifeform which uses the body of Nick Gillespie as its host). While a location in Santa Clarita or Mission Viejo would be more in keeping with the Cox/O’Toole values often espoused by the publication, it’s also a bear of a commute for the 25-35 year old who just isn’t interested in living in those places, period. And so it comes to pass that Reason Magazine is now located across the street from a bus layover point.

Kind of sweet, really.