U.S., Andrew Selden and Amtrak: Spend Less, Get More; Use $66 Billion Wisely For Network Growth Avoiding Mistakes Of The Past And Squandering It On Random Political Demands

By Andrew Selden, Guest Commentator; November 16, 2021

Investment Opportunities

How can Amtrak make the most use out of $66 billion in cost–free new capital?

Congress in its role as Amtrak’s investment banker has seen fit to spend as much as $66 billion over ten years on a vague notion of non-commuter “rail passenger service.” While more than half of that is already earmarked through simple political power (rather than objective investment criteria) to further enhancements of the Northeast Corridor (NEC), the legislation is conspicuously lacking in criteria for allocating the remainder.

Because Amtrak’s national market share for intercity passenger transport is trivially small (motorcycles produce more annual passenger miles nationally; in the NEC, buses carry more passengers than Amtrak’s trains, leaving Amtrak with an NEC intercity market share of perhaps 1 ½%), one might conclude that the upside for new investment into intercity rail is quite large. But that still begs the obvious and vital question: where and how can the new capital not earmarked for the NEC be invested in a way that maximizes the return on the investment of scarce public capital?

The alternative is to squander the money on random uses dictated by ordinary political processes rather than applying objective investment criteria.

Before doling out any of this new money, the Federal Railroad Administration (FRA) should ask itself, “Where and how can we apply the available new capital in such a manner as to drive the greatest growth in output (measured in annual revenue passenger miles (RPMs)) per dollar invested?” Anything other than that amounts to frittering away cash and opportunity.

Amtrak, blithely assuming that all the money is for its use, has a scheme called “Connect U.S.” that is a warmed-over version of its 1973 “Emerging Corridors” effort to draw more subsidy than it was already getting from states outside the NEC. Connect U.S. is a scheme to develop dispersed and discontiguous short interurban corridors, featuring frequent, fast trains, and modeled largely upon the NEC.

The NEC and the other regional corridors (funded by states, outside the NEC) are Amtrak’s smallest and weakest business segments by all relevant objective criteria such as intercity ridership, transportation output (RPMs), load factors, market share and return on capital investment. FRA embraced the underpinning of Connect U.S., ignoring its conflict of interest as being both the steward of taxpayers’ capital and the de facto owner of Amtrak.

The Amtrak/FRA scheme will necessarily minimize the return on investment of the new public capital. In addition, its opportunity cost (forfeited opportunities elsewhere) is vast.

If the goal is to maximize the return on the new capital, short stand-alone corridors are not the answer.

Short Corridor Costs

In all public transport, it is a tautology that frequencies must be inversely proportional to distance in order to attract sufficient patronage as to warrant the effort. The shorter the distance, the greater the frequencies must be. Once-a-day service might work in 2000-mile transcontinental corridors, but not in urban transit markets or 200-mile intercity markets.

But … greater frequencies in rail service quickly require greater investment in infrastructure. Higher operating speeds exacerbate the infrastructure challenges, and operating frequent, fast passenger trains in a corridor owned and used by freight carriers quickly becomes highly demanding on infrastructure, and extremely expensive to mitigate.

Examples abound; here are two: 1) The infrastructure challenges to absorb rapidly growing demand in the Pacific Southwest Coast corridor have stifled growth there for decades, and, 2)

The costly infrastructure used in the NEC to accommodate a Byzantine mix of traffic ranging from transit “stopping services” to transient inter-regional trains to pretty-fast Acela trains, all on the same railroad, not only costs around a billion and a half federal subsidy dollars a year, but also incurs another half billion a year (more or less) in deferred maintenance and delayed investment that today has accumulated to a fixed facility deficit of well over $30 billion.

New corridors won’t be any easier, illustrated by the current catfight over Amtrak’s apparent national priority of adding two trains a day between New Orleans and Mobile, a $100 million investment that is projected to lose another $7 million a year once open. That is not a good return on investment.

Wherever Amtrak looks to develop a new short stand-alone corridor featuring frequent, fast trains sharing mainline railroad space with whatever freight traffic exists, infrastructure investment will be required in fairly substantial amounts, for passing sidings, incremental main tracks, PTC in places, signal and communication systems, stations and station pull-outs, grade crossing abatement, etc.

And in most of these markets, including the NEC as well as the Gulf Coast, the investment doesn’t yield much. In almost every existing market in the U.S., these short corridors suffer from half (or more) empty (short) trains, low output of RPMs, high operating costs, and embarrassingly low market share.

Perhaps the greatest cost of these new corridors is the opportunity cost of much higher-yielding investment opportunities forgone by the misallocation of finite capital.

In short, the business model envisioned by FRA and Connect U.S. boils down to: “Spend more, get less.”

Return on Investment

But what if the advent of the new money from congress, whatever is left over after the majority of it is spent a priori in the NEC, prompts FRA to invest differently, in a way that is consciously designed to maximize the return on investment by maximizing the growth in annual RPMs per dollar spent? What if we instead decide to “Spend less, but get more”?

Instead of modeling growth with the new money on Amtrak’s smallest, most overcapitalized and commercially weakest markets, perhaps the emphasis should shift to building upon and replicating Amtrak’s largest, most undercapitalized and strongest markets. The inter-regional (long distance) trains are already Amtrak’s largest and most successful, and under-capitalized, sector measured by intercity ridership, annual RPMs, load factor, market share and return on invested capital.

Load factor is the key to guide new investment. Load factor is the percentage of annual available seat miles that are converted into revenue passenger miles by being sold to paying passengers. Where capacity consistently exceeds demand, the chronic low load factor means that the sector is already over-capitalized. In these markets, Amtrak is wasting capital by producing more inventory than can be sold.

Where demand consistently exceeds capacity, on the other hand, the high load factor means that sector is under-capitalized. In these markets, demand is being turned away for want of carrying capacity. Applying new capital here will add capacity to accommodate demand, and raise output. This approach maximizes the output of new annual RPMs per dollar invested.


The purpose of congress’s new spending is to increase the use of intercity rail passenger service.

Two different kinds of growth exist: organic and scale. Organic growth means increasing the annual sales of existing units or channels. Scale growth means adding outlets or channels. They are not mutually exclusive, but pursuing scale growth makes little sense where organic growth lags. Different investment strategies are involved to attain each kind of growth.

Imagine a small chain of fast food stores. Efforts to increase annual sales at existing units are aimed at organic growth; adding locations is scale growth. But suppose that most of the existing units are under-performing, with weak sales and poor returns on invested capital. What owner would borrow new money to build more losing outlets? What banker would fund it? Yet that is exactly the foundation of the FRA/Amtrak growth plan for the new capital soon to be available from congress, America’s most indiscriminate lender. A priori assumptions contradicted by consistent empirical experience are a poor guide to new investment.

If the goal is to maximize growth per dollar invested, the focus instead should be to direct that capital to scale growth in the markets that are objectively undercapitalized, while bringing more non-capital efforts (price reductions, focused marketing, promotional incentives) to bear on achieving organic growth in the under-performing markets to sell inventory currently going unsold. At some point, if those efforts falter, capital will have to be withdrawn from over-capitalized sectors (and re-directed to undercapitalized sectors) and capacity reduced better to align with actual demand. Replicating these under-performing services in new markets is a seriously bad idea.

The Role of Networks

If attention is directed to the goal of maximizing growth in output per dollar spent, the effects of networks must be considered.

In any network, output is primarily a function of the number of nodes in the network; in transport, this is the number of usable origin/destination pairs in the network. (To illustrate the concept here, we can disregard differences in sizes of the nodes, as they all average out over the network, especially in larger applications.) The number of accessible O/D pairs in a network is a proxy for its output.

Additive expansions to the network drive exponential growth in the network, because each single addition interacts with all other nodes. The arithmetic expression for this is that the output is equal to the number of nodes (stations), times the number of stations, less the number of stations (because no single journey starts and ends at the same station; a round trip is two one-way journeys), or: N squared minus N (N(2)-N).

These network effects can also be observed in other systems: e.g., telephone grids, the internet, urban freeway grids and airline hub-and-spoke systems.

Disconnected short corridors involve small values of N, and so while they add modest output, it can never approach the yield in increased output from extending and interconnecting longer routes with many more stations into a more effectively-integrated national network where the value of N is large to begin with and grows significantly with additive expansion. The squaring function of the equation assures that.

Extending and interconnecting long corridors has the lowest capital cost (new infrastructure needs are modest), by wide margins the greatest number of resulting O/D pairs in the network, and fits well within the infrastructure and operating needs of host railroads, with comparatively minor enhancements. Spend less, get more. The return on investment in incremental annual RPMs per dollar invested is of an entirely different scale than with isolated short corridors.

Network Opportunities

Four examples illustrate the concept.

  1. Building a smaller-scale regional network: Amtrak sends two trains a day from New York to Eastern Canada (Covid aside), the Maple Leaf to Toronto via Albany and Buffalo, and the Adirondack to Montreal via Albany. These trains do not interact except by doubling frequencies between New York and Albany (and Schenectady). One of them (which one probably doesn’t matter) instead should originate at Boston, on a schedule designed to bring the two into Albany at the same times each day for a cross-platform “hub,” with the dwell calculated to maximize the probability of making connections every day. The incremental cost is near zero, as the distance from Boston to Albany is only slightly greater than from New York. The N value of the network equation expands enormously because it is now possible to get to or from everywhere south and east of Albany and everywhere north and west on the affected routes. And major new markets are opened up, e.g., Boston-Montreal, for free. Spend less, get more.
  2. Route extension: Each year (Covid aside), about 4 ½ million Canadians, most from the Toronto and Montreal regions, visit Florida. Amtrak’s share of that market is zero because it has lacked the vision to tap into it. The solution is to extend the Silver Star (because it, unlike the Silver Meteor, serves Tampa; a majority of Canadian snowbirds are destined for Florida’s west coast) to Toronto and Montreal, with the train splitting into separate sections at Albany, just as the Lakeshore Limited and the Empire Builder split at Albany and Spokane, respectively. This changes the Star into a two-night train like most of the western inter-regional trains. The cost of one or two additional trainsets, labor and track rent is modest compared to the billion or more in infrastructure and equipment investment needed for many of the short isolated corridors. The value of N in the network equation for this change is huge, and it opens major new markets and a huge existing traffic flow. Spend less, get more.
  3. National network growth:  The existing Chicago-Oakland Central Transcontinental Corridor can be expanded significantly with an extension of a single existing daily regional train, and only 175 new route miles (each way), by extending an existing Missouri River Runner beyond Kansas City to Omaha, and running it through to and from Chicago in place of an existing Chicago-St. Louis train. This opens up for the first time for Amtrak all the potential traffic from populations east of Omaha in Kansas City, middle Missouri, St. Louis and downstate Illinois to and from Omaha and points west served by the California Zephyr. Do people in St. Louis travel to Denver, Reno or Northern California? Of course, but Amtrak’s share of that is zero because they don’t run a connecting train. Once again, modest costs, a huge N value, and potentially huge new traffic. Spend less, get more.
  4. More national network growth: All four of Amtrak’s western transcontinental routes can be handily interconnected, opening up significant new markets, and enabling for the first time easy connections among them, by installing a new train connecting St. Paul and Dallas, via Des Moines, Kansas City, Newton (Kansas), and Oklahoma City. This sort of re-creates the Rock Island’s Twin Star Rocket, and would be more expensive than other similar network-building expansions, so it might not be a first-tier effort. This train would leave St. Paul southbound after 9:00 AM, leave Kansas City shortly after arrival there of the Southwest Chief, and arrive in Oklahoma City in time to run on the current schedule of the Heartland Flyer, and vice versa. Again, the costs are real but relatively modest — far less than creating a stand-alone high-frequency, higher speed short corridor — and the yield potentially enormous. The purpose is much less to carry traffic locally between stations on the one new route as to open up overhead opportunities in the very large matrix of O/D pairs accessible by use of the new service to interconnect four transcontinental routes. Spend less, get more.

In 1985, United Rail Passenger Alliance studied making similar modest changes to the Chicago-Los Angeles Southwest Transcontinental Corridor. This study was reported fully in the January, 1986 issue of Trains magazine. A rigorously calibrated gravity model was constructed for the route. The model was asked what the results in traffic would be if three changes were made to the route structure: (i) split off through cars at Barstow for San Jose via Fresno and Oakland (displacing a local train already running between Bakersfield and the Bay Area, so the only new train miles are Barstow-Bakersfield); (ii) drop through cars at Flagstaff for Phoenix and Tucson; and (iii) split the eastbound train at Kansas City, with a section running to Chicago via St. Louis (in place of a local train already running, hence no new train miles). The model inherently assumed no change and no improvement in the nature or quality of the train service being provided, i.e., do exactly what the trains were already doing, just do it in a larger matrix of O/D pairs.

The model said that the average daily on-board count on the eastbound train between Flagstaff and Albuquerque would jump from 212 to around 1300. THAT is real growth at modest cost, maximizing the return on whatever the modest investment would be. All of that could be done for the cost of one low-yielding short corridor. Spend less, get more.


The once-in-a-generation allocation to passenger rail of $66 billion in cost-free capital gives rise to a one-time opportunity to allocate and prioritize the uses of that capital by rational, objective standards based on getting the biggest possible bang for the new bucks. The only way to do that is by seeking investments that maximize the number of new annual RPMs that can be generated by each dollar spent.

The ratio of capital cost to output and the synergy of networks tell us that stand-alone short corridors with frequent, fast, short distance trains absolutely minimize the ROI on invested capital, while expanding and better integrating longer routes absolutely maximizes the amount of passenger service that can be generated by the new capital.

FRA can repeat the errors of the past, or it can chart a new and far more productive course, based on how it chooses to allocate the new money. The new money also means that with just one or two years appropriations, FRA can test the hypothesis by setting up an open and transparent demonstration project: one stand-alone short corridor (e.g., Phoenix-Tucson, or New Orleans-Mobile) and one long route integration (e.g., the Silver Star to Canada, or the Kansas City-Omaha connection).

The results of that experiment would then guide future prioritization of use of the new capital. The alternatives, repeating the failures of the past, or allowing parochial politics to determine the priorities, would be a tragic waste of money and opportunity.

Please share with others