IFP’s Transit Abundance Playbook
Two weeks ago, the Institute for Progress published a playbook called Transit Abundance, comprising 15 policy briefs about reducing construction costs. One of them was my own, about itemization of costs. Eric has a good one about phasing funding to prevent the situation of the Boston Green Line Extension in which the planning for the project was funded from the same grant that funded construction, leading to rushed pre-planning to get that money in the first place. And unfortunately, many of the rest are questionable. There’s a mix of good, meh, and downright counterproductive, and unfortunately, I’m seeing people I know and respect from elsewhere say things that are wrong or barely relevant.
My first filter for pieces like this is “is this even close to correct?”. There are some pieces that fail the test right there, praising projects that were extremely expensive as examples to emulate rather than avoid (for example, the reliably incompetent Stephanie Pollack is saying that of the Green Line Extension).
But beyond that, for things that at least get what the problem is, I look at what they cite. If they only cite the Transit Costs Project, then they don’t know enough to play. If they cite things I wasn’t aware of, then it’s more interesting. A positive highlight is Anthony Potts’ piece calling for technical scoring of proposals, which relies on literature about the exact weighting required for technical scoring to matter. We knew from interviews that California’s 30% technical score, 70% cost weighting produced identical results to lowest-bid. This piece actually went and found published results on this issue and found that if the cost is weighted at 57% or more then it’s the same as lowest-bid, and that this produces higher final costs both in general and in California case studies. This piece, I encourage everyone to read for a more complete discussion than we wrote of technical scoring and its merits.
And if this piece is a highlight, a lowlight is a piece on AI, by Lizzie Speed and Bennett Capozzi, panned across social media for calling for using more AI for transit planning, from a background of trying to sell AI technology rather than from that of transit planners; the authors cite the Transit Costs Project’s review of planning and other soft costs to argue the unique Anglosphere problem is data limitation (we say no such thing in our report) for which the solution is more AI, and seem not to know where those high costs come from. This compounds a criticism I’ve seen across social media for IFP’s use of generated images, to the point that the lead image for the playbook was noted for having errors in how the trains look (it’s hung on bus-like wheels rather than on bogies).
And regrettably, I’d like to also highlight Jackson Moore-Otto’s piece, which says that federal loans can stabilize transit funding to prevent a feast-or-famine problem. I respect Jackson and have seen him do and lead good work on matters of commuter rail and Boston-area advocacy, and am honestly surprised he’d write such a thing, which is just wrong.
The specific problem with Jackon’s argument is that it relies on a chain of citations, each introducing some errors to the point that the final claims are just wrong. A central claim is that the first electrification project costs more than subsequent ones, which is manifestly untrue (for example, compare Amtrak’s Northeast Corridor electrification in the 1990s with its struggles to install autotensioners now). This is based on a claim sourced to Momentum (which doesn’t quite say that), which sources it to the RIA Electrification Cost Challenge (which gets some Continental costs wrong); to get there, Jackson is saying the New Haven-Boston electrification was not a first line but a subsequent line, when in fact Amtrak had literally never electrified a railroad before and its forerunners on the Northeast Corridor had last done so in the Great Depression. The British costs are meanwhile compromised by repeated project cancellations that lead to sunk planning costs and make internal comparisons difficult, since the surviving lines are by definition the easiest ones. In truth, rather rapid electrification projects, without decade-by-decade continuity, exist outside the UK and are often quite reasonable in their costs, as in Denmark, Israel, and especially India.
The upshot is that because there is no feast-or-famine problem with electrification, any program to reduce its cost in the US (or UK) needs to understand what works and not talk about making it easier to apply for federal loans. On the Northeast Corridor, for example, a senior engineer since removed for incompetence and potential embezzlement of Amtrak funds unilaterally decided to tighten the required catenary pole spacing, driving up costs for reelectrification since the existing poles are not compatible with the new standard (but are compatible with high-speed electrification systems sold by international vendors). This is not a feast-or-famine problem, and to the extent there may be a specific federal policy solution, it’s to require planners to attend international industry fairs and conferences like InnoTrans and make it clear that ones who don’t will have stalled careers.
Pedestrian Observations 
Would a reasonable argument instead be that rolling programs may make it more likely that standards are adopted? And that rather than the rolling program being the root cause of low costs, it’s rather an ecosystem in which standards are not being constantly relitigated?
In particular, I am thinking of the PRC system in which HSR planning, design, and approval can happen in less than a year — all under CRC’s umbrella. Or Spain, where ADIF AV does the same but slower AFAIK (e.g. ~10 years for Madrid-Galicia). Moving from HSR to metro, again PRC has clear standards while the degree of customisation seen in the US (Los Angeles Sepulveda line’s expansive range of heavy rail options) and UK (Elizabeth line plug PSDs) is high.
Is it possible to quantify the degree to which standards are used in transit planning?
Well, PRC HSR construction costs relative to the amount of tunneling aren’t particularly low. France’s are lower; Germany’s are inconsistent, depending on whether money is wasted on excessive tunneling or on other things.
Standardized systems are important, and there is a systems premium in the US and UK, but it can’t be about a regular pipeline because London has that and yet Crossrail engaged in some really dodgy systems decisions that drove costs up.
Thanks.
Is that also correcting for % viaduct? In my experience, PRC HSR has high % viaduct as well as high % tunnel, compared to European HSR. I don’t have numbers, and that anecdata will be skewed by mostly taking PRC HSR in the mountainous South.
I am also thinking that a rolling program would bias you towards good standards, but that certainly is not an iron law. It could be for instance, that the PRC has fallen into a local minimum in the parameter space, given that there are only a few design agencies able to bid for HSR and metro (AFAIK, ~5). An interesting analysis would be to see where each country is most constrained in terms of capacity.
No, if you correct for viaducts then China is cheaper. But that raises the question of why China uses viaducts so much whereas here we’re fine putting HSR on the ground on earthworks when there’s no tunneling.
I wonder if it’s an engineering practice learned from Japan. As in Japan, elevated structures are very common in PRC not just for HSR but also for auto expressways and metros. Cities in other Asian countries that also may have learned such from Japan: Taipei, Manila, Bangkok, Jakarta…
The reason Japan does it is that when you have weak eminent domain powers, fragmented landownership plus lots of earthquakes, typhoons, floods, landslides etc, the net cost of a viaduct is not so bad. That’s not a bad fit for Korea/Taiwan which have similar land law and similar restrictions.
Last major earthwork burst was in the early 1960’s with the Tokaido Shinkansen, 5 Direction JNR project and the Musashino line. But the surface area of embankments is higher than viaducts, and actually the concrete used is increasingly similar, all of these earthworks are encased in concrete slaps as reinforcement. 1970’s lines from Sanyo Shinkansen to the New Town Lines
Notably Japan is also allergic to trenches not just embankment earthworks/at grade. Soil liquification is a thing.
Doesn’t mean its a good thing to do on the North China Plain…
China also adopts ridiculously conservative gradient and curve radii standards for HSR, with an all EMU fleet as well. That probably makes grade changes much more difficult and makes an all-viaduct alignment easier to design (while also forcing more tunnels). Taking into account all of that China probably isn’t that more expensive.
Thanks for the reference to the Transit Abundance playbook. The “Eliminate Redundant Subway Cross-Passages” article by Brian Potter is very interesting.
The National Fire Protection Association (NFPA) 800 feet cross-passage requirements accepted by the Federal Transit Association are stricter than the 500 m European standard and Japanese expert opinion that these standards are unwarranted. The NFPA had admitted to a “lack of technical substantiation” for its requirement.
The transit agencies can either follow the NFPA standard and have a better chance of securing federal financing while inflating construction costs, or they can perform an engineering analysis demonstrating that wider spacing will result in an acceptable level of safety and ask for an exemption. Brian Potter notes that “this places a significant burden on the transit project designing team, as it requires them to perform an analysis and risk failing to secure approval, possibly delaying the project”.
That transit agencies need to perform an engineering analysis to be exempted from a standard written without engineering analysis suggests that US transit practice not only suffers from its isolationism from foreign achievements but tends to isolate itself from empiric experience in the first place. I have seen the same thing in medicine with practice standards set out of experts guesswork instead of experimental data.
We’re looking at cross-passages and their costs and so far everything looks 2.5th-3rd order.
What is the sum of US standards issues? To include: over-customisation (quantifiable?), onerous environmental review for obviously environmentally beneficial projects (if we may lump this in as a standards issue), poor standards (per Alain), and standards fragmentation (e.g. elevator codes)? I would guess that the sum of those is an order of magnitude more than just cross passages.
Interestingly, the FY2024 FRA awards for the NEC showed quite a difference in electrification cost per mile if the scope of the project was limited to catenary structures and wire versus a more comprehensive replacement of the traction power system, including substations, transformers and grid tie-in; and of course a comparison to Northend is imperfect because Northend did not need to decomission and remove existing traction power infrastructure and maintain electric services during construction. These numbers omit state matching funds:
I’d be interested to know how electrification cost scales with number of tracks. It stands to reason that traction power for 4 tracks is not 2x the cost of 2 tracks, but by actually how much, I can’t tell immediately.
The electrification numbers I use always include transformers and substations. If anything, this should reduce the cost of installing autotensioners, which does not require new substations and transformers.
Electrification costs are supposed to scale with the number of tracks, as far as the wires go; substations depend on peak traffic but are degressive (for example, we have cost estimates for expanding the Boston-Providence substation to accommodate MBTA electrification and they’re very low).
Thank you for your reply.
The fact that the incremental cost of implementing electrification for at least a couple of MBTA commuter rail lines is within the realm of the achievable makes this nonsense all the more infuriating (see slide 4, although you’re probably already well aware).
Back of the envelope math: electrifying 9 miles of Fairmount at $11M/mile, ~$100M for a small fleet of multilevel EMUs (based on NJT), ~$50-75M for any high-level station mods required and ~$50-75M for a layover yard/maintenance facility, plus whatever you’ve got for Boston-Providence gets you a starter pack of EMUs and partial electric service on two lines, somewhere in the neighborhood of $300-350M. Instead, they’re paying $200M over that, for the privilege of renting battery consists for 15 years.
It seems the high level platform modifications on the line have exploded in price for some reason. It seems that MBTA is estimating 55-90 million per station to be upgraded for full accessibility, and they’re currently proposing “temporary” mini highs for 3 million per station that will probably need replacing every 10 years. Using this cost estimate the two stations without full high platforms probably could be upgraded with <150 million dollars, even at the egregious cost estimates.
The MBTA needs to be able to upgrade stations much more cheaply going forward however, as the whole system really should have full high platforms.
$11M per mile for dual track catenary is also on the high side. I picked it because it’s Northend escalated for inflation, give or take. I wanted to provide a reasonable comparison to a capital project for electric services.
As with all transit costs, we should endeavor to frame our arguments around what things should cost in a transparent world, not what some consultant spun out of whole cloth.
The messaging about these BEMUs, their costs, and capabilities are rooted in untruths and incompetence. I am reminded of Alon’s post about Connecticut’s procurement of unpowered coaches.