In 2008, Congress required that railroads install positive train control, which would automatically cause trains to slow or stop to prevent derailments or collisions, on all lines that carry passengers or hazardous materials by December 2015. That deadline is two years passed, yet–as last week’s accident revealed–still has not been met by most railroads.
The Washington train wreck was a special case. The rail line, improvements, passenger train, and upgrades were owned or done by four different government agencies. It seems particularly galling that neither Sound Transit, which owns the tracks and is spending billions on rail construction, nor the Washington State Department of Transportation, which received close to a billion dollars from the federal government to upgrade this particular line, bothered to install a working version of positive train control before inaugurating service on this route.
In general, however, the railroads have two very good reasons for not enthusiastically installing positive train control as Congress has demanded. First, the cost is high: the Federal Railroad Administration estimates it will cost as much as $24 billion, which is probably more than the annual capital budgets of all the private railroads in the country.
The second reason the railroads are delaying is that the benefits of positive train control–or at least the version that the railroads are implementing–are trivial. Take a look at this record of transportation fatalities. While about 35,000 people died on highways in 2015, only about 750 were killed in railroad accidents. Of those 750 deaths, positive train control will significantly reduce only those in the first line–“train accidents” as opposed to “grade crossings” or “trespassing.” Over the past ten years, the average of the train accidents line is just 10.
Compare positive train control with driverless cars. All of the hardware for driverless cars will be built into the cars themselves, adding perhaps $1,000 to the cost of a new car. By comparison, positive train control will require both modification of locomotives and considerable new infrastructure on the railroads. So the costs are much higher.
Driverless vehicle technology may reduce auto fatalities by as much as 90 percent–say, 30,000 a year. But even if it only reduces them by 50 percent, driverless technology will produce lots of other benefits: less congestion; increased productivity as people can spend travel time doing more productive things; fuel savings; and so forth. Positive train control, however, won’t eliminate the need for someone to drive each train. The only practical benefit is that some fraction of 10 lives would be saved each year.
Every accidental death is a tragedy, but to minimize those tragedies we need to carefully spend our limited resources. I suspect the railroads could save a lot more lives for a lot less money by fencing rights of way and thereby reducing the number of trespassers killed by trains.
As a result, the Federal Railroad Administration (FRA) itself admitted in 2009 that the cost of positive train control “would far exceed the benefits.” But it is still required to make sure the railroads follow the law even if the FRA thinks the law is a waste.
It didn’t have to be this way. As an early FRA paper observed, there are two ways of implementing positive train control: either as an overlay on top of existing railroad signaling systems or as a replacement for those systems. Transportation engineer Steven Ditmeyer points out, if used as a replacement for existing systems, positive train control can produce “shorter train running times, improved running time reliability, improved track capacity, and improved asset utilization.” When all of those benefits are counted, they far exceed the costs.
So why aren’t railroads doing it that way? Ditmeyer’s short answer is that, when Congress created the mandate, the railroads gave the problem to their signaling departments rather than their information technology departments. The former saw positive train control as something to be added to what they were already doing while the latter would have seen it as a replacement. Ditmeyer’s longer answer was that the railroads should have hired “system integrators,” but I suspect it also has something to do with the internal politics of the railroads, the FRA, and anyone’s natural resistance to being ordered around.
All three of the people killed in last week’s Amtrak accident were rail fans eager to take a trip on a newly-opened route. As a rail fan myself, I sympathize with them and their families and feel some outrage that the government agencies that owned and operated the track and trains couldn’t be bothered to comply with the law two years past the deadline.
At the same time, the train had another system on board designed to prevent accidents, one that has proven itself to be pretty reliable after nearly 200 years worth of experience: the engineer. Why this particular engineer failed to slow down at the 30 mph signpost has yet to be revealed, but I’m not willing to blame an expensive and unimplemented technology when something else was clearly at fault.