Short trains and crashing the boards

If you want railroads to improve safety, service and sustainability; stop pushing for policies that will hurt rail service, safety and sustainability.

Those reading who have ever watched modern professional basketball with someone who used to play or coach at an amateur level, have likely heard them complain about the players being lazy and not going for rebounds. While watching the USA vs Serbia semifinals game during this year's Olympics in a hotel lobby, an older person at the table next to me was loudly complaining that the US players were loafing, and that they needed to crash the boards to prevent Serbia's NBA MVP Nikola Jokic from grabbing all of the rebounds. He thought the US was behind by up to 17 points and going to lose because our players lacked energy and didn't care about collecting boards.

He was wrong, the US didn't adjust their rebounding strategy, and they ended up coming back for the win and the gold medal.

Modern players purposefully avoid crashing the boards not because they're lazy, but because it's bad strategy. Any wing who runs in to grab an offensive rebound is a wing who isn't running back on defense to prevent a transition basket in the other direction, and has a low chance of generating a board against well positioned centers like Jokic. There are exceptions to this; above the break 3's bounce harder off the rim and are easier for wings to catch, but in general NBA coaches have rightly followed the lead of the legendary Gregg Popovic who would instruct his players that "We don't care if you get an offensive rebound in your entire life."

The video below shows what happened on one play when players didn't run back on defense before a Nikola Jokic rebound. The result was a catastrophic and violently quick dunk in the other direction.

How does this relate to rail?

Speaking very generally, for the last couple decades, North American freight rail seems to have been divided between two very different orientations. For simplicity, let's call them revenue maximization and operating ratio strategies. Revenue maximizing strategies are generally more popular with freight customers, union employees, sales teams and the public. Operating ratio strategies have been most popular with executive teams and investors, exemplified by Precision Scheduled Railroading as a specific example of an operating ratio strategy.

Revenue maximizing strategies are typified by

• Safety first
• Expanding business and routes
• Lots of R&D and capital spending
• High employee head count
• Careful maintenance of equipment
• Priority trains-Amtrak or preferred trains that travel at higher speeds and for which slow moving bulk trains ahead of them will stop at sidings to allow passage
• More and shorter trains, with maximum train length rules
• Running all locomotives on trains and pulling locomotives out of storage to increase horsepower

Operating ratio strategies are typified by

• Safety fourth (per Hunter Harrison's ordering of priorities)
• Eliminating underused tracks, train symbols and hump yards
• Reductions to R&D and capital spending
• Heavy cuts to employee head count
• Loose maintenance of equipment
• Reducing train passes for priority
• Fewer and longer trains
• Shutting off excess locomotives on trains beyond minimum needed for grade
• Moving as many locomotives as possible into shortage

The revenue maximizers are right on the generalities, but catastrophically wrong on the specifics

Here are the things that revenue maximizers get right.

• Safety always needs to be a railroad's first priority. Nothing is typically more important to an individual than staying alive and healthy, and as such railroads have a responsibility to preserve their employees and the general public's health.
• It makes no sense to track chiefly the operating ratio instead of total profit, as shareholders returns are based only on the latter. Additionally, a focus with blinders for only a lower operating ratio has historically been associated with lower capital spending leading to reduced returns in the long run.
• Freight railroads consume only about a quarter of the fuel per ton-mile that trucks do. This lead is falling as the trucking industry improves its efficiency rapidly with help from the DOE's Supertruck program and independent efforts from corporations along with drivers, but remains substantial. Trains also have only 1/8th the fatalities per ton-mile compared to trucks, saving the lives of many on highways and residential streets. Thus railroads are both climate and safety tech, and have a duty to expand their business from their current 28% of the freight market in the US, for the sake of both railroads shareholders and the world. As such, the past few decades have not been the time to cut on trackage, yards, personnel or capital spending!
• Customer service is critical and railroads have historically done a very bad job of it. Providing decent ETA's and GPS tracking of railcars and containers is especially important and underfunded.
• Railroads underspend on maintenance. Wheels, bearings and air brake systems are often maintained only to minimize replacement and repair costs under FRA rules, when replacements and repairs should be happening much earlier as poorly performing parts will harm fuel efficiency and safety well before breaking FRA limits. Tangent warning: this is especially true for bad wheels which add rolling resistance (and for which FRA rules on raw Max Kips are a terrible measure as it is influenced by train speed and car weight among other variables when the rule should be based off of (max kips)/Avg(kips) or (max kips)-Avg(kips); for air brake leaks which increase fuel consumption and dwell time as locomotives require more time and fuel to pressurize the system; and for rail lubrication where locomotive equipped friction modifiers are severely underutilized.

Where revenue maximizers go off the rails is when they start complaining about longer trains, locomotive storage plus shutoff rules, and train passes for priority. On these topics critics suffer from the same issue as basketball fans who complain about a lack of offensive rebounding effort - their heart is in the right place wanting their team to win or railroads to offer better safety, service and sustainability. They believe that the main actor isn't doing what they want only because of laziness or a deranged focus on just the stock price and operating ratio (a fear that freight railroads have done a lot to earn.) But in both cases the protagonist is doing the right thing, despite the fact that it looks awful, because its the better strategy on almost every level. As NBA players should usually avoid running in for rebounds as it will leave them open to transition baskets in the opposite direction, so railroads should run longer trains (even longer than the ones they are running post PSR), keep locomotive utilization low, and avoid train passes for priority. All three of these policies help achieve safety, service and sustainability, despite uninformed or biased claims to the contrary.

With freight trains - Bigger is better

Over the past year there has been a lot of public discussion around the benefits and risks to the public of longer trains. It started with a John Oliver episode where he complained that larger trains block crossings for longer and can increase derailment risks. The key scene has a Thomas the tank engine character forced to pull a comically stretched out train by a maniacal Operating Ratio focused executive, resulting in a derailment and explosion. It continued with a Transportation Research Board paper on longer trains from September 17th of this year that had a mostly negative take towards their impact on blocked crossings, derailment risks and Amtrak trains sharing the same tracks. The AAR came out with a letter the same day disputing this and making the point that

"An arbitrary limit on train length would risk disrupting the nation’s supply chain through increased network congestion and undermining environmental priorities. For example, AAR analysis found that restricting train length to 7,500 feet would increase U.S. mainline freight train fuel consumption by 13% or an additional 423 million gallons of fuel annually. This would be the equivalent to the annual emissions from about 930,000 cars."

A single gallon of diesel fuel produces about 22.4 pounds of CO2 (considerably more than the weight of the fuel itself as it reacts with Oxygen from the atmosphere.) The lowest EPA estimate for the social cost of a ton of carbon is $120 so a 7,500 foot rule would cause ~$516 million of environmental damages annually before considering separate increases in NOx and particulate emissions. This would obviously be a catastrophe, and fits entirely with my expectations from years of running regressions on freight train fuel data, where train tonnage was always the first variable that had to be accounted for as a confounder. Longer trains are more fuel efficient because

1. They have fewer locomotives per ton which forces the locomotives they have into higher throttle notches where they are more efficient
2. They have lower aerodynamic drag per ton as a train has only one head end with no drafting benefits on the first locomotive and only one rear wake

However all three of these sources are neglecting the single most important safety question with regard to longer trains. Their impact on grade crossing and trespasser collisions.

In 2023 96.7% of US non-suicide rail fatalities came from collisions that occurred with either trespassers or at grade crossings according to the FRA. This included 247 deaths at highway-rail crossings and 715 trespasser deaths. There is some evidence that some suicides are being misclassified as accidents but these numbers remain very high regardless. The vast majority of these collisions occur on the head locomotive and as such the frequency of these events should be almost totally independent of train length. By increasing size, train count should fall by proportionally the same amount, reducing the total number of accidents.

If we were to reverse the gains in train length since 2010 of ~19%, we would expect ~132 extra non-suicide deaths a year that aren't suicides in the US. This is assuming that non-suicide collision deaths are proportional to train count and that trespasser deaths are being overcounted due to missed suicides by 37.5%. This adjustment for missed suicides makes the estimate of lives saved smaller and more conservative.

The benefits from saving ~132 lives a year are immense and hard to comprehend. It feels a bit improper to put a dollar figure to it, but the US Department of Transportation puts the value of a single life saved at $13.2 million. This means the lived saved by increasing train lengths in the US since 2010 are probably worth somewhere in the neighborhood of $1.7 billion a year.

This puts the total benefits from longer trains in the US to the public due to just lower carbon emissions and fewer grade and trespasser collisions at around ~$2.2 billion a year.

It is certainly true that a train of 150 cars will almost always be more dangerous than one of 75, be more susceptible to break in two events due to higher buff and draft forces, unable to use shorter sidings, and will take around twice the time to pass through a grade crossing. But this is never the choice that railroads face. 150 cars will have to be transported on either a single train, on two trains, or on one train and ~75 trucks. As long as doubling train length produces less then a doubling in safety risks per train it will be the safest option overall.

Notes: Doubling train length producing less than a doubling in safety risks per train seems to be the case on every measure with the exception of break in two risks. Break in two's are almost certainly better managed through proper train makeup rules including mandatory DPU on longer trains and avoiding consists that mix together Autoracks with cushioning and manifest cars. As break in two's have vastly lower fatality risks then grade crossing or trespasser collisions choosing a tradeoff of fewer grade crossing accidents for a few more break in two's is an extremely easy decision.

On the subject of blocked crossings, there doesn't seem to be any evidence that this is getting worse beyond anecdata. As there is a gap in time between when crossing gates close and open before and after a train the same amount of railcars on fewer trains should mean that crossings will be closed for less time in total. Additionally the first second that a car is stopped is by far the most costly, due the fuel and time costs of braking and accelerating. So if there is a tradeoff between stopping a car for 10 minutes with 1 train or 2 cars for 5 minutes each with 2 trains, it is generally better to stop the single car. Lots of other blocked crossing complaints are about network fluidity and claims about seeing more trains stopped on the crossings, as longer trains mean fewer train passes they typically mean better fluidity and fewer stopped trains though many other PSR type policies can hurt fluidity (too few engineers, conductors, mechanics etc.)

Railroads operate like assembly lines, not highways

When a truck driver is traveling at 55 mph in the right lane, it is easy as a driver to get into the left hand passing lane, briefly accelerate, and then get beyond them. Outside of rare situations where crew districts have triple track or double track with no trains coming in the opposite direction, this doesn't occur in rail.

Instead when a faster train with higher priority overtakes a slower one, as often happens with passenger trains on freight networks or premium intermodal trains, the slower train must slowly brake to stop at a siding, allow the faster train to pass, then slowly accelerate back to full speed afterwards. As proper blocking distances must be maintained between trains they will be stopped for much longer then the direct time it takes for the faster train to pass them. While they are occupying the siding it can't be used to allow trains in the other direction to come through or for any other purpose. By the time they have made it back to full speed they will have fallen far behind their original plan and risk slowing down more trains behind them. They may be forced to stop at another siding.

Per FRA regulations train crews are limited to 14 hours of work before a train must be stopped even if it has not made it to the next yard. The slowest trains on a network are the most vulnerable to being stopped for priority passes, and the most at risk of a recrew event that can occur if they have to stop one too many times. When this happens they must stop at a siding and wait for a van to pick them up. The train will remain at the siding, or even the mainline if there wasn't a siding in the area to use, clogging up traffic, until a new train crew arrives to pick it up. Sometimes the congestion from chains of events like these gets so bad that one train will require multiple recrews just to get through one "crew district."

When a network gets clogged like this, every train using the network is harmed. In this way, strictly enforced priority passing rules for passenger and premium intermodal trains will often reduce the velocities of even the passenger and premium intermodal trains themselves.

This chart contained in a simulation paper by Mussanov, Nishio and Dick from the University of Illinois illustrates the point very well. The red lines show train delay compared to plan when passing rules for priority are implemented and the yellow lines show total delay when all trains are given equal priority. The dotted lines are for non-priority trains (flexible) and the regular lines are for high priority trains (scheduled.) Both red lines are higher than both yellow lines showing that both types of trains deteriorate in performance when passing for priority is implemented. Giving the scheduled trains enforced passing for priority rules SLOWED THEM DOWN because it destroyed the network's fluidity. The same can happen when the government or rail customers enforces passing for priority rules for their own trains. The x axis on the chart shows that usually (but not always) scheduled trains are better then flexible departure ones for network fluidity.

In general, rail networks should be viewed more like assembly lines then highways when considering impacts on velocity. Reducing train speed variance on a line, passes and stops is usually more important then increasing raw speed. Increasing the speed of the trains that are already the fastest may actually damage network velocity. Can you imagine how hard Charlie Chaplin would have it in the below gif if these objects were moving at different speeds and he had to pull some off the line to allow others through?

If you see Thomas the Tank Engine on the rail, shut him off

In Q1 2018 Union Pacific was struggling with its velocity. Carload growth, a hurricane and other issues had combined to harm average train speeds which fell from 25.7 mph in Q1 2017 to 25.1 mph in Q4 2017, then down again to 24.8 in Q1 2018. A number of initiatives were tried to improve velocity.

Key among these was an attempt to bring more locomotives out of storage and to pause locomotive shutdown rules. Train engineers are normally ordered to shutoff or idle locomotives in excess beyond those needed to meet minimum "horsepower per ton" or "equivalent powered axle" per ton rules that railroads make. These rules typically differ between train categories and routes to make sure that all trains can make the maximum elevation climb on a route and sometimes to give premium trains a horsepower bonus.

Railroads record engineer compliance to these rules to save fuel as using excess locomotives increases fuel burn due to the fact that locomotives are generally more fuel efficient in higher throttle notches. When outputting the same amount of power from two locomotives set on throttle 3, vs one locomotive set on throttle 6, the single locomotive will likely burn much less fuel. Additionally locomotive shutoff rules help reduce locomotive failure events for the same reason that longer trains help to reduce accidents. Though a single locomotive in T6 will have a higher failure rate then a single locomotive in T3, it will generally have a lower failure rate then two locomotives in T3. Compliance to these rules does not generally harm velocity as

1. The number of locomotives needed to make maximum grade is usually enough to also make the speed limit on the majority of the route or at even grade
2. Reducing the number of excess locomotives also reduces variability in train speeds, which helps to reduce train passes and stops, thereby improving fluidity

Putting extra locomotives into storage that would otherwise end up as excess locomotives on trains further helps as it

• Reduces the extra tonnage and length created by unused locomotives on trains
• Removes the least fuel efficient, most polluting, most likely to have a mechanical failure older locomotives from the system
• Can reduce yard dwell related to maintaining, fueling and switching excess locomotives

Of course if a railroad does not have enough locomotives in the system to meet minimum horsepower per ton rules velocity and fluidity will be harmed as trains will be stuck in yards waiting for locomotives. But the relationship between locomotive count in the system and network velocity is not monotonic it's parabolic, just as too few locomotives can harm speeds, so can too many.

Searching for any way out of its velocity issues in 2018, Union Pacific tried out what seemed like an easy solution at the time, just using as many locomotives as it could for about 3 months in Q1 and Q2. The results of this experiment can be seen in the table below taken from UP's quarterly financial results where C-rate is gallons per thousand ton miles.

Union Pacific burned an extra 33 million gallons of fuel compared to what would have been expected under 2017 efficiencies. Fuel burn continued to be elevated in Q3 as the excess locomotives were slowly taken out of the system and engineer shutdown compliance took time to reset. The social cost of carbon was similar to UP's $ fuel spend at ~$40.7 million. Velocity was not helped by the extra locomotives and continued to fall, all the way to 24 mph in Q3 2018.

Pulling locomotives from storage continues to be a solution that I hear for railroads struggling with velocity, but it is not one that works outside of extreme cases where carloads have really jumped. If you see Thomas the tank engine on the rail shut him off, as steam locomotives along with very old EPA Tier 0 diesel electric locomotives should not be running trains in 2024.

A middle way for rail

To state the obvious, instead of working to only maximize revenues or minimize operating ratios, railroads should work to maximize long term profits. To me this means

• Safety first.
• Expanding business and routes-Moving more freight from highways to rail is critical for profits, the environment and to reduce traffic deaths.
• Lots of R&D and capital spending-Freight rail is nowhere near where it has the potential of going. Imagine autonomous battery electric railcars like those being developed by Parallel rail and Intramotev among others. They will offer much more then just autonomy and a switch from Diesel to electric. Much quicker and safer braking with power regeneration, perfect railcar tracking and health maintenance, incredibly rapid switching, better track usage with smaller signal blocks and much greater flexibility in train size. Not just on the low end with the ability to run single car consists, on the high end as well with the outright elimination of buff and draft forces. Trains will still benefit from size increases to improve aerodynamics and to reduce crossing accidents. Aerodynamics at the railcar level can also improve with technologies like the ones my company Deflect are discovering, as can rolling friction, maintenance and customer service with technologies other companies are developing. If the rail industry refuses to improve at all it will be leapfrogged and destroyed by trucking which is exploring trucks with vastly improved aero, battery electric systems and autonomy. Rolling friction will always be better with rail but its other advantages may disappear if the industry slacks on technological improvement.
• High employee head count-railroads are currently understaffed, and increased short term profits from layoffs hide long term losses from future customer sales, technological improvement and mechanical failures that change as a result.
• Careful maintenance of equipment-just following FRA rules isn't sufficient as those rules lack good subjectivity and sensitivity for predicting accidents and malfunctioning equipment can create efficiency and safety issues long before failure.
• Train passes for priority don't work. They belong to a strain of win-lose thought which holds that priority trains can win only by stopping non-priority ones, but both train types lose in the end as network fluidity is destroyed.
• Longer trains are safer, more sustainable and better for network fluidity/service. The most important safety questions are reducing road-rail crossing deaths, trespasser deaths and moving more freight from trucking to rail. Longer trains help with all of these. They also reduce fuel consumption and help network velocity by reducing train count and congestion. Where in the past railroads have built bigger trains by combining train symbols and consolidating trains, they should instead create bigger trains in the future through increased business catalyzed from improved customer service and performance.
• Locomotive count at both the train and network level must be rightsized. Too many locomotives will hurt fuel efficiency by raising the horsepower to ton ratio and worsening the average EPA tier of locomotives in service; but can also paradoxically hurt network velocity by increasing locomotive failures, train passes from increased speed variability, and all of the yard work that locomotives create. Network velocity will be damaged if locomotive count is wrong in either direction.