The high cost of Amtrak accounting
by Andrew C. Selden and E. P. Hamilton III
Reprinted with permission from Passenger Train Journal, 1984.
- Part I
- Amtrak’s Accounting System
- Some Business Basics; A Word on Fixed Costs
- Resource Allocation Theory
- Operating at the Margin: What it Means
- Part II
- The RPS System
- The Folly of Averages
- Fuel
- Food Service
- Reservations Service
- Equipment utilization and scheduling
- Part III
- The station closure program
- Conclusions
Andrew Selden is a partner in a Minneapolis law firm, practicing in the area of business and commercial law. He has published several papers and lectured on the subject of the economics of intercity rail passenger services, and Amtrak accounting and planning.
E. P. Hamilton III received his Ph. D. in Engineering from the University of Texas at Austin, where he conducted research in optimization methods for cost control and marginal cost planning. He is Chairman of the Amtrak Committee, Austin Chamber of Commerce.
PART I
A farmer, impressed with statements on the need to cut transportation costs. decided to buy no more feed for his horse. This was the only form of transportation he had, A few weeks later a neighbor asked him how the experiment was working. ‘It started off just fine,’ he said, ‘but just as the old horse was getting used to the idea, it died.’ — Rad Latimer, CN Rail, quoted in RAILWAY AGE
A business lives or dies by its accounting system as much as by what it does in the marketplace. This is as true for Amtrak as for any other business, because the accounting system is the manager’s window on the world. Unless the accounting system delivers a current, complete and accurate picture of the financial condition of each element of the business, the manager is literally flying blind. His business can fail before he knows it and for reasons he couldn’t suspect.
In these articles, we will explore Amtrak’s chronic failure to take proper account of fundamental economic and business factors in their route accounting and planning system, resulting in major planning and marketing blunders and the loss of numerous opportunities to expand services throughout the nation on an economically sound basis. improving its revenue-to-cost ratio and reducing its operating deficit.
As early as 1975, the late Joseph V. McDonald, an Amtrak board member, discovered Amtrak’s route accounting was arbitrarily charging train (not service) crew costs to the Montrealer at a rate sufficient to pay for 26 enginemen and trainmen per crew, rather than the actual complement of five men. He concluded that Amtrak route accounting was arbitrary and misleading. The same thing is still going on in 1984.
As we will see. Amtrak’s accounting system may overstate the projected cost of adding or retaining any particular train by as much as 1700 percent. Coupled with management’s persistent sociometric bias in favor of certain specific Northeast Corridor (NEC) submarkets, the accounting system’s distortions cost the nation hundreds of millions of dollars annually in wasted costs and lost opportunities. The accounting system thus precludes informed route planning and evaluation, and causes huge overcharges to 403(b) states. Amtrak’s system conceals the profitability, at the margin, of many carefully selected new rail passenger markets.
These failures have nothing to do with the accuracy of Amtrak’s annual financial statements, which reflect the overall condition of the corporation. The problem, rather, is how Amtrak internally compiles and redivides costs, charges costs to trains, and projects costs in ways that would bankrupt a private sector business or, as with the short-sighted farmer, starve the company’s horse to death.
Amtrak’s Accounting System
The core of Amtrak’s accounting system is the Route Profitability System or RPS. Amtrak describes RPS as its only route accounting system, where all cost and revenue data from thousands of reporting points are collected, classified and reassigned to particular routes or trains, to overhead or to other operating accounts.
RPS is the source of much of Amtrak’s purported short-term avoidable cost data. but RPS cannot and does not account for the actual costs of individual routes, trains or facilities. Rather, it collects costs into categories, then reallocates those costs to various facilities and activities in accordance with arbitrary formulas based upon management’s assumptions and preconceptions. In many cases, readily available empirical data is not used in allocation formulas. Graham Claytor characterized RPS as a “. . . fully allocated route accounting system which allocates costs to trains.” at about the same time that Amtrak’s senior planning officer described Amtrak’s “short term avoidable costs,” derived from RPS, as “. . . costs that in our judgment are truly incremental to a particular service.” A system that allocates costs cannot produce true marginal costs.
The crucial importance of understanding RPS lies in its uses by Amtrak. Multiple copies of RPS cost reports are routinely circulated to senior marketing and planning officers, presumably for decision-making purposes, despite its admittedly fully-allocated nature. Fully-allocated cost data is not useful for most decisions. Amtrak executives, however, rely on RPS data and reports for route planning, evaluation, 403(b) costing and billing and reports to outside agencies. This data is then used both within Amtrak and in DOT, OMB and Congress to make critical decisions on which trains and routes live and which die.
RPS data is also used to estimate costs of proposed 403(b) services, and to bill costs to states that are parties to 403(b) service contracts. Because the cost allocation formulas are based on management assumptions rather than facts and empirical evidence, management can manipulate data and formulas to support decisions management may make impulsively (such as rerouting trains 5 and 6 to Colorado) or arbitrarily (such as discontinuing the Spirit of California or implementing such cost hemorrhages as the New England Metroliners).
Amtrak has also admitted publicly that RPS is biased in favor of the NEC and against national system trains, and that RPS-based Route Profitability Reports, widely circulated within Amtrak and Congress, are “useless and misleading.” Route Profitability Reports are often the only economic data many members of Congress see, regarding Amtrak’s performance, and accordingly are the only basis on which they can make decisions on Amtrak appropriations and route eliminations.
In the second part of this article, we will examine case studies of how Amtrak route accounting works in certain specific cost categories, and its effect on specific planning issues, to document that the system is useless to the purposes it serves. Before doing so, however, we must explain some basic business economics.
Some Business Basics
Economics is sometimes described as the “dismal science,” often rightly so. Rail passenger advocates, nevertheless, must understand how successful businesses are managed, what their data requirements are, and how proper planning occurs. to comprehend the magnitude of Amtrak’s route accounting and planning errors. Even the most public-service-oriented advocates will agree that Amtrak has a mandate to spend tax money in the most prudent fashion, maximizing profit, nationwide levels of service, and overall return on investment, from each dollar invested. Maximizing profits and maximizing output at any given investment level are two sides of the same coin.
Economics textbooks place considerable emphasis on how costs and revenues operate together to allow managers to choose output or activity levels that maximize profits. Costs cannot be analyzed without considering effects on revenues. Revenue opportunities always have associated costs. called “opportunity costs,” and costs cannot be arbitrarily and independently controlled without a potentially adverse impact on revenue and profit. Analysis of marginal costs and revenues allows the manager to evaluate how the interaction works in specific cases.
Marginal costs are sometimes called “incremental” costs. The terms are synonymous. Marginal costs are the actual costs of the next unit of activity in a business. the next product produced or the next unit of service rendered. A marginal cost may include incremental capital outlays as well as increased variable costs. Marginal cost analysis should be applied to both proposed increases and reductions in output. A marginal cost is always a real, actual, traceable cost of doing something. Marginal costs and revenues are almost always nonlinear. See fig. 1.
 |
Two approaches to handling data: A nonlinear cost curve represents the actual costs of a business such as a train. This curve is simplified here but will probably have discontinuities as well (fig. 5). In contrast, the straight line is a linear “approximation” based on averaged data.Notice that
The straight line is what Congress was led to believe would happen if a route were cut or added, even though the late Amtrak board member Joseph V. McDonald showed that over 75 percent of a Amtrak’s fixed costs charged to a single route (fixed costs are the majority, according to the ICC) remain if a route were cut, and are simply redistributed to other lines. Amtrak’s financial department reluctantly agreed with him. |
An “average variable cost,” on the other hand, is almost never anything more than an arithmetic mean that has no value for planning or evaluative purposes. An average variable cost is nothing but the sum of an activity’s variable costs divided by the total number of units of output.To illustrate, the average variable cost of a hamburger shop is the total cost of its meat patties, buns, wrappers, condiments, etc., divided by the total number of hamburgers cooked. This average value tells the shop manager absolutely nothing useful about the actual cost or profit of producing one more, or less, hamburger in the next accounting period, or whether he made a profit in the last period, or what price he should charge for his hamburgers. Average costs and revenues, like any average value, almost always vary linearly.
Non-linearity (see fig. 1) is a natural effect of the phenomenon of an economy (or diseconomy) of scale, which reflects fixed investment seen as over or under capacity. The available capacity, in turn, influences production cost and selling price of goods and services. Confusing marginal with average costs can be disastrous:
I make widgets. My accounting system calculates. allocates and projects costs based on averages. My market is not price sensitive so I can sell all my widgets. regardless of quantity. at $2 each, My plant is not running at full capacity. I rent the plant and machinery on a short-term lease, so I have no fixed costs to speak of.
Last year I made 200 widgets at a total cost of $150, sold them for $400 and netted $250. My average cost (“avoidable cost,” if I am considering reducing output) per widget according to my accountant was $.75 ($150 / 200). I can make widgets only in batches of 100, and my operating budget (the most my banker will lend me) is $200. Should I increase my output next year to 300, or 4OO? My accounting system tells me no, because the projected cost of producing 300 widgets, $225 ($.75 average cost x 300), exceeds my budget. Accordingly, I decide to stay at 200 units and pursue cost control initiatives such as laying off my sales agents.
My nephew, an industrial engineering student, studies my operation for a class project. He notes my underutilization of plant capacity: He does a cost study and discovers that, in fact, my actual cost is not a uniform $.75 per widget, but $1.00 per widget for the first batch of 100, $.50 each for the second batch, and the next 200 would really cost only $.25each. He writes a term paper on the classical economy of scale available in my plant and recommends that I make more widgets because the profit on the next batch would be $1.75 each, compared to only $1.00 each on the first batch.
Did my average-based accounting system and my linear cost projection (after all, my accounting system shows that my costs vary linearly) serve me well? Or did it cost me a lost profit and my customers an extra batch of widgets, if only I had invested an extra $50 in costs?
I lost big. My 200 widgets last year, cost $150: $100 for the first 100. $50 for the next batch (not a uniform $.75 apiece). But the next two batches would only have cost $25 each. or $50 total. So I could have made 200 more widgets, and grossed $800 by selling 400 widgets, at a total cost of $200 (my maximum production budget). My profit would then have been $600n enough to finance three years’ operations or expand my output for still greater profit.
Output Cost Revenue Profit Average Cost Method: 200 $150 $400 $250 Marginal Cost Method: 400 $200 $800 $600 My incremental profit alone from using the correct accounting and projecting method is 140 percent greater than the total profit from my incorrect average-based method. See fig. 2.
Figure 2 |
This curve represents profits (or also total revenues) and illustrates a problem common to organizations which use averages because they do not take enough of the right data.To reach maximum profit and the optimal level of output at C, should output be increased or decreased?
Notice that management is flying blind – there are no numerical values on the axis representing service output, because insufficient amounts of available data are collected. Suppose management does know tthe present profit or loss (level 3 in this case). That value cuts across the curve at A and B. At point A, output must increase to increase profit; at point B it must decrease. Without correct data, there is little chance of making the right decision. In reality the situation is much worse: the lack of good data makes the size and shape of the curve unknown. The answer: obtain better data and use it properly; fig. 4 shows how. For a business to operate correctly, one must know the correct relationship of output, profit, price and cost. |
Almost all business decisions are made (in successful businesses, at least) “at the margin.” based on an evaluation of the actual cost and potential profit of a change in level of output, not of average or total costs at either level of output. That evaluation requires an accurate assessment of the true marginal cost of the proposed change, as well as the impact of the change on revenue and profit.
Because the marginal cost of a proposed change is its actual cost, successful planning cannot be done on the basis of any average costs. not even the average variable costs, of existing levels of output. A cost is not a marginal cost unless it is caused by and traceable directly to the change being evaluated. An average cost, such as an average variable cost (or a “common variable” cost, one of Amtrak’s unusual classifications). may vary generally with widely varying levels of output, but unless the planner can identify a direct causal relationship between the change under evaluation and a specific, identifiable cost, no marginal cost is involved. See sidebar.
The level of output that ordinarily yields maximum profit is that volume whose marginal cost equals the price at which the output can be sold. This is known as volume pricing at the margin. The same principle also determines optimization of level of output for a given subsidy or deficit level, using the same analytical technique. See fig. 3 and fig. 4 for a graphical representation of a simple profit-making business.
Figure 3The folly of averages illustrated. |
Cost-cutting using averaged data by itself will not work. Attempting to minimize the short-term avoidable cost (which is actually the average variable cost obtained by averaging and ratio-ing) results in bankruptcy – shutdown.Even minimizing fully allocated costs (averaged total cost) will not generate a profit – it only results in the business breaking even.
The correct strategy is to use actual (not averaged or ratio-ed) marginal costs, targeting a volume otuput so they exactly equal marginal revenues. The correct method is not a simple cost-cutting limited-growth philosophy: cost cutting is a politically palatable government exercise, but it simply will not work in a business and it will not work at Amtrak. |
 |
| Here is the answer to the dilemma of fig. 2.These two graphs show the complex interaction of marginal revenues, marginal costs, and total revenues as they relate to output.
Consideration of the entire family of curves shows whether to increase or decrease service. More importantly, the marginal cost shows the effects of scale which can be exploited to maximize profits. Notice the difference between the marginal cost, total cost, and total profit curves. The way they look shows an important rule: You must spend money to make money. The converse of this rule is that cutting costs alone will not maximize profits. |
Correct use of the marginal cost method automatically recognizes the nonlinearities caused by the effects of scale and is applicable to a wide variety of demand (price and market) conditions. Such nonlinearities are readily apparent even in empirically-derived curves such as Fred Wengenroth’s passenger train cost curves (September 1983 PTJ), reproduced in fig. 5. Figure 6, however, shows the corresponding marginal cost curve derived from his data. Compare the curves. The downward-sloping sections of the marginal cost curve represent an economy of scale, caused by the fact (as Wengenroth recognized) that once a coach is added to a train, it costs essentially the same to operate whether anyone rides in it or not. The same is true in many other aspects of any transportation business. including station operations.
 |
Fred Wengenroth’s empirically derived total cost curve for the Black Hawk (September 1983 PTJ), including estimated crew deadhead costs, shows how the passenger train cost function an become nonlinear even with a minimal amount of data. Notice that the curve (left) is a series of steps, not a straight line at all. Mr. Wengenroth also points out some of the shortcomings of Amtrak’s data base from which the curve was derived, and how the states with 403(b) service are pressing Amtrak to provide more realistic cost data. |
Figure 6. |
The marginal cost curve corresponding to the nonlinear curve in fig. 5. Several facts are apparent:
A linear “apprximation” of costs based on averages (or, the average variable cost or short-term avoidable cost) will wipe out the above curve, replace the useful with significant misrepresentations of the actual situation, and lead one to believe incorrectly that costs increase in ratio with the amount of service offered. |
Average variable costs, on the other hand, eliminate and conceal all effects of non-linearity necessary for correct analysis. The effect can be disastrous, as in the case of the widget factory. Indeed. it can be shown mathematically that attempting to control costs using average data, ignoring revenues and nonlinearity, will inevitably call for major reductions in output, resulting eventually in a level of operation sufficient only to finance a shutdown of the system. This means bankruptcy for a real business, and shows up at Amtrak as reduced long-haul services, inadequate consists, chronically understocked and understaffed food service cars, inadequate and misplaced corridor services and continuous and growing deficits.
Unlike most private businesses. Amtrak does not employ the cost classifications of “fixed” (those that do not vary with levels of output), “semifixed” (that vary, but not directly, with changes in output) and “variable” (that vary directly with output). In our hypothetical hamburger restaurant. the rent and property taxes are fixed costs. labor and electricity are semifixed, and the cost of the meat patties and buns is variable.
Amtrak instead calls its costs “fixed,” “long-term avoidable,” and “short-term avoidable.” Short-term avoidable costs include both “direct” costs of train operations. such as fuel (whose cost is allocated. not measured). track-usage payments to contracting railroads, and engine crews, and “common variable” costs, such as arbitrary percentages of commissary, reservation center and equipment maintenance costs.
In Amtrak’s system, a short-term avoidable cost is one that ends when a train is discontinued. Note the curious regressive management outlook this nomenclature reveals. A long-term avoidable cost is one that varies with levels of train operation, but is not traceable directly to a particular train. Fixed costs exist regardless of train operations. In PY 1981, Amtrak’s fixed costs for the NEC alone were $280 million; its fixed costs for the rest of the national system were $126 million.
Amtrak uses the STAC (Short Term Avoidable Cost) category, derived from RPS, as its main planning tool, and as the basis for 403(b) billings. This all sounds very rational, until one looks more closely, at what this system produces. Long-term avoidable costs for specific trains often are calculated not independently but as an arithmetic ratio to total nationwide costs in various categories. STAC is not an expression of any given train’s own costs, but mostly an extension to individual trains of shares of system-wide costs in various categories based on arbitrary linear formulas, such as car-miles or passenger boardings. The formulas, in turn, are ordinarily based on management’s assumptions about the character. source, relationships and variability of costs, rather than on actual causality or real train-specific data. If Amtrak were running our hypothetical hamburger shop as part of a chain, its STAC calculation for the one shop would not be that unit’s own costs of its hamburger patties, buns, etc., but an allocation to that shop by a formula concocted from management assumptions of an arbitrary share of the entire chain’s costs for meat, buns, etc. In fact, this is exactly how Amtrak accounts for its food service operations.
Avoidable costs are always charged against train revenues, actual or projected, to determine profit or loss; fixed costs are not, except in the reports Amtrak sends to Congress. For example, Amtrak makes NEC trains look profitable in comparison to national system trains by treating identical costs as fixed in the NEC and variable elsewhere, based on management’s bias and political assumptions about the sanctity of NEC services. rather than the inherent nature. variability or source of cost.
Amtrak’s practice of averaging system cost category aggregates also has a pronounced tendency to spread much higher costs of high speed, high frequency, high employee-passenger ratio, premium corridor services in the Northeast out to other lower-cost operations in the NEC and elsewhere. Matching costs to the activities that cause them has little or no role in Amtrak’s accounting system, even for many easily measurable direct costs of train operations.
A Word on Fixed CostsKnowing the level of fixed and semifixed costs in different segments of a business is as crucial as knowing the actual marginal costs of a proposed activity. Another basic element of successful business management is that the greatest marginal profit opportunity lies in that portion of the business with the highest ratio of variable to fixed costs. This is because in those segments more of each dollar of revenue remains available, after paying direct and variable costs of operations, to amortize fixed costs and flow to the bottom line as profit. Amtrak’s long-distance trains not only have by a wide margin the greatest revenues in the system, by train and by segment, but also the very highest ratio of variable to fixed costs. Thus, it is in this segment of Amtrak’s business that the greatest opportunity lies to earn incremental operating revenue and profit-to help amortize the system’s huge fixed costs. It must never be forgotten that fully 55 to 60 percent of Amtrak’s ticket revenues of its passengers account for 75 percent of its passenger miles. Amtrak does not lose money on these trains as trains; the high-revenue Western trains all cover their operating costs from their own ticket revenues. It is only the failure of these trains’ revenues to cover Amtrak’s allocated fixed and semifixed costs, 69 percent of which come just from the NEC, that produces the deficit. |
Now here is the rub: Amtrak’s planners claim their short-term avoidable cost category is the same as a business variable category, and, worse, that the RPS-derived short-term avoidable cost of any given train, existing or proposed, is also its marginal cost. In fact, Amtrak’s short-term avoidable cost in most cases is really nothing more than an average variable cost figure.
As we will see in the second part of this article, this mistake is extremely costly. What compounds this error is that Amtrak’s planners, in estimating the cost of a proposed new train, will extend the average variable cost of existing trains to the proposed new service on a linear basis, which’ totally ignores and denies any possibility of an economy of scale resulting from increasing use of underutilized common and fixed facilities. Amtrak does this by using linear extension formulas, such as “car miles” or “passenger boardings,” to extend system averages to new services on a straight line basis. Their accounting system, after all, shows that costs vary linearly. But marginal costs are not average costs-the seemingly linear behavior of Amtrak’s variable costs is produced by its average-based accounting system, not real-world causality. Average values are inherently linearly variable. That is why they cannot be substituted for actual marginal costs in route accounting or planning.
Amtrak’s average-based systems conceal that, in many markets, the real cost of a new train would be substantially less than the average cost of existing trains. This was the case with the Spirit of California, which Amtrak accounting concealed by arbitrarily assigning shares of system expenses that were not. in fact, caused by the Spirit or directly variable with its operation. This accounting method is also how Amtrak buries the huge marginal maintenance-of-way costs caused by 120 mph Metroliner operations. Amtrak allocates MOW costs to its own trains in the NEC (which exceeded $80 million in FY 1982) by “unit miles” rather than speed-factored gross ton-miles. The huge marginal MOW costs necessary to operate Metroliners at 120 mph are allocated not just to the 120 mph trains that cause the cost, but to all trains using the line, including 79 mph long hauls and through corridor services. This adversely impacts all trains including non-Metroliner NEC services.
Resource Allocation Theory
There are both profitable and unprofitable, counterproductive, ways to run a business, and to monitor and manage its costs. One way to reduce costs is to cease operating. Costs will then approach zero. This approach, however, does little for the revenue side of the profit equation. Another non-optimal way is to emulate the farmer–cut costs to the bone, even if the output then available drops to zero. Like unmanning or bypassing stations, this approach cuts costs all right. but the horse may die in the process. It does eliminate the nuisance of having to service customers. In the private sector, bankruptcy is the inevitable result. Even budget airlines staff their stations.
The decision process which analyzes costs and revenues to maximize profit is called resource allocation theory, the assignment of resources (e.g., money) to various activities to maximize profits. The theory requires a thorough understanding of one’s own business and marketplace. Powerful mathematical tools are available to help in the allocation of resources. Amtrak programs can be analyzed using these tools to determine if they were good or bad business decisions. Such an analysis reveals an incredible degree of ineptitude by Amtrak management. Later, we will apply these principles to Amtrak’s planning and route accounting processes to illustrate how badly flawed they are.
A fundamental tenet. of successful business planning is calculating your return on investment, either projected for a proposed new activity, or actual from an activity being conducted. Without that data, a business cannot know where to invest, or whether to invest, its capital resources to make the highest return, or indeed whether it would be better off buying a CD at its bank. Incredibly, Amtrak does not or know its rate of return on investment in its various trains, routes or promotional activities, and thus cannot allocate its limited resources in a rational fashion.
Unless capital is freely available, a business certainly could not borrow or raise equity capital without knowing and showing the investor that the capital could be deployed to yield a return to the investor and a profit to the business. The manager or planner cannot decide rationally how to deploy .finite resources to competing demands unless he knows which will give the highest_ return. Amtrak does not have this information and seemingly lacks the ability or awareness to calculate it. Both the planning and marketing departments have denied responsibility for projecting or achieving even a positive rate of return from major investments of Amtrak’s extremely limited capital.. This failure helps to explain why endless millions of dollars can be poured into saturated NEC markets, where the returns to date have been negative, while other lucrative markets are ignored.
Amtrak’s notorious New England Metroliners. the fourth consecutive failure of limited-top express service in the U.S. in the last 15 years. achieved a 50 + percent reduction in ridership and a net loss of revenue despite extremely high fares, free food, extremely high employee-passenger ratios (easily twice that of a typical long haul), legrest coaches and a $1,000,000 train-specific ad campaign.
How could Amtrak have made the decision to invest millions in a high cost service that would carry only 40 to 50 people per trip, and yield less than 4 percent of the number of revenue passenger miles as a typical Western long haul? Graham Claytor admitted he did not even know whether the fare premium covered even the cost of the service amenities. We know it didn’t cover the cost of the ad campaign. Such hideously wasteful investments (those dollars could have gone instead to support neglected but still ultra high revenue services in the West and Sunbelt. for example) reflect the failure of Amtrak’s route accounting and planning systems, and its incomprehension of elementary business investment strategy. It also explains why Graham Claytor thinks there is no prospect for growth in the national system in the next decade.
Amtrak’s internal route accounting systems, therefore, do not properly reflect basic principles of business economics. They grossly distort the concept of marginal costs and normal business planning methods and goals. They are manipulative. As a result, Amtrak’s resource allocation (i.e., route and service and capital investment) decisions are based on impulse, political pressure and sociological theory rather than business objectives of maximizing profits and return-on investment, and optimizing levels of output. In the next installment, we will analyze several case studies that illustrate and document these failures.
Operating at the Margin: What it Means
Dr. Paul Samuelson’s definitive treatise on economics notes that the fundamental principle upon which business is based, operation at the margin, is sometimes hard for the layman to comprehend, but is actually easy once one realizes that, in economics and business, the tail wags the dog. The success or failure of a business is based largely on how its managers produce and react to change. A change can affect total profit or loss disproportionately to the size of the change itself.
Common sense tells us that any decision, including a business decision, is actually a determination to change something. For example, if I decide to quit my job and start my own business, I am weighing the costs and benefits of making that change in my career. I project what I will make by changing course, and what it will cost. I make the same analysis (albeit more routinely) every time I weigh whether to change the output of my business, even by as little as one item. I weigh the costs and benefits of the change itself, not its global effects based on some total or average. I do not have to develop a complete profit/loss sheet just to look at the effects of the change — indeed, that would be costly, time-consuming and misleading. I cannot use averages because they do not reflect what is actually occurring when I make the change.
Suppose you are hungry for soup, but you don’t have any. At the store, your favorite is priced at $.59 per can, or 2 for $1. You went to buy one can, but you know you will want more later; should you change your plan and buy that second can? What does the change cost? Well, I can buy two cans for a dollar, so the average price for two is $.50. This is my average variable cost. (The average variable cost for one is $.59.) But, how much does that second can, the change in my plans, cost? $.59? Certainly not — that’s the cost of the first can. The average variable cost of $.50? Not at all. You can’t buy any soup at that price. Nor can you add the average to what you planned to spend (the original $.59), because that total is $1.09 and we know the total for two cans is $1.00. Using the average or average variable value predicts a price that is almost 10 percent higher than what you actually will be charged. Extending the average variable cost for one ($.59) on a linear basis (per can), which is how Amtrak projects costs for new trains, predicts that the total cost will be $1.18, fully 18 percent more than the actual cost. In fact, the correct answer is that the second can costs just $.41, and if you change your original plan and buy that second can, $.41 is the cost to you of making that change.
This principle applies in the business world as well. Suppose I own a manufacturing plant. I will choose the level of output that will bring me the maximum profit (not unit sales or gross revenue) in a given marketplace. I make that choice by considering the cost to manufacture the next item I produce or forego producing (i.e., what does it cost to change my present output by one) and price I could get for it. If the price I can get for the next item exceeds its cost, I will make a profit on it, so it pays to produce it. Then I consider the next item, and I continue until I no longer make a profit for the next item. At that point I have made the most profit I can, so I hold output constant until some other factor changes. I did not use aggregates, averages or any other shortcuts to come to the conclusion above. I used only real data, at the margin.
The actual cost of making that next item is the marginal, or incremental cost, and the price I can get for it (with some subtleties beyond the scope of this discussion) can be called the marginal, or incremental revenue. The marginal cost of the first can of soup is $.59, and the second just $.41. In the manufacturing example above, I maximized profit by increasing output until the marginal cost equaled the marginal revenue. This is called volume pricing by marginal costs. One variation of it is the way the food store priced the soup to induce you to buy more than one can. Note that the average variable cost, which Amtrak uses to make (or justify) its decisions, is misleading and dangerously counterproductive as a decision-making tool both to the seller and buyer.
This method, when used properly, has a twofold effect on profits, a “double whammy” if you will. By recognizing the dynamics of supply and demand, it allows you to price to include volume sales (incentive pricing) while simultaneously producing the optimal volume of goods or services to meet demand at that price while maximizing profits. It is, by nature, and an expansionary philosophy, although it also provides the correct analytical approach toward cutbacks. In practice, the individual costs are known, but are grouped together as nonlinear mathematical functions (averages are not used) and analyzed as a whole. This allows computation of the marginal cost and revenue directly at any given level of output, rather than by going through the tedious one-at-a-time exercise used above to illustrate the concept. Methods used generally are based on an extension of Lagrangian multiplier theory and are part of a specialized branch of mathematics known as the calculus of variations. Other accepted optimization techniques, such as nonlinear or dynamic programming, can also be used. Methods not generally applicable in a large or successful business include the four-function calculator, aggregate averages, or the back of an envelope. While the whole concept may seem quite complex, almost like a magic wand which optimizes profits, it is well-grounded in mathematics because it requires understanding of the dynamics of the marketplace and the complex interaction of costs, revenue and volume.
The problem at Amtrak is that all too often its executives project the cost of its next unit of output by calculating its current average cost per car mile or train mile to be x, then concluding that the next car mile or train mile will also cost x. This conclusion is rarely, if ever, true. You can’t buy any can of our hypothetical soup for the average variable cost of two cans.
PART II
“[Marginal cost-based pricing] unfortunately leads in the railroad business to an attempt to obtain profitability by generating high volumes of traffic, all priced on a strictly incremental basis.” (Emphasis added.)– Letter from W. Graham Claytor Jr., March 25, 1983
Here are some recent consequences of Amtrak’s average-based planning and accounting systems:
- During the 1983 Christmas traffic peak, Amtrak’s headquarters terminal. Washington, D.C., ran out of diesel fuel, delaying fully loaded long-haul trains up to six hours.
- Over a 23-month period. Missouri Pacific overcharged Amtrak for Eagle train crews in a single crew district by an average of $34,700 per month. This was not detected by Amtrak, but by a MoPac employee.
- Amtrak routinely closes stations, citing lack of funds, while it pays laid-off station agents full salary under the C-2 program. Management sees these as unrelated issues.
- Amtrak laid off an entire maintenance staff at full pay under C-2, then 150 miles away paid a contractor to maintain a single train, and hired a second railroad crew to maintain another single train on an adjacent track at the same time. refusing and covering up all requests by the laid-off personnel to relocate if they could get their jobs back.
- An Amtrak station agent in Arkansas quadrupled his station’s business with a privately funded local advertising program, only to be investigated by local management under orders to find a reason to fire him.
- Amtrak assigned two F40 locomotives to 4-car New England Metroliners, but only one F40 to the 10-car Superliner Eagle, ordering crews to lash the engine’s rear door open with a coat hanger to enhance air flow over the overloaded generator to prevent fires. Several Eagle engines failed en route. One caught fire, almost destroying a million-dollar locomotive, delaying the train eight hours, and requiring that it be towed with a full load of high-revenue passengers but without head end power for several hundred miles.
- Amtrak cites insufficient equipment to meet demand or expand service yet sells cars as often as possible; the last offering included over 110 Budd stainless-steel cars which can be converted to HEP for a fraction of the cost of the new cars being prototype. Amtrak threw away the Superliner tooling.
- The Eagle was delayed for an hour while its train chief argued with management by long-distance phone to add a coach sitting idle at that station, to accommodate 24 angry long-haul standees, while management balked over the cost to run the car.
These incidents provide alarming insight into Amtrak management’s naive views of economic and even simple operational issues discussed in the first part of this article. which can severely affect the economic viability of a route or even the entire national system. Failure to understand or observe such fundamental concepts as pricing at the margin and maximizing the return on investment, when coupled with these other types of decisions, leads to the notion that something is very seriously amiss at 400 North Capitol. What Graham Claytor views as “unfortunate” is, in fact, generally agreed by successful business managers and economists to be optimal in the business world. Amtrak’s failures in these areas are a major cause of Mr. Claytor’s no growth policy, the station closure program, triweekly service, the preoccupation with cost containment at the expense of revenue generation, a pervading emphasis on low revenue, high cost, saturated short-haul markets, deteriorating relationships with 403(b) states, and failure to utilize the car fleet effectively.
The extent of Amtrak’s misunderstanding is not confined to the national system. While it has caused numerous regional inequities, even the NEC is not immune to its negative effects. In the ICC’s findings regarding Amtrak’s costing methodology for the Northeast Corridor, Ex Parte 417, it commented that Amtrak had made a “misinterpretation of the concept of avoidable costs as it relates to the sharing of common costs.” The ICC explained, using detailed examples, how economies of scale work and how actual marginal costs-not average costs as Amtrak was proposing-had to be used if an accurate statement of costs, fixed and variable, was to be developed. The ICC made clear that a realistic marginal cost method was not now in use even in the NEC. There is a very clear opportunity to restructure even the NEC, with its glut of costly services operating beyond the point of diminishing return, to provide a higher quality of more profitable services to more people in more Northeastern markets, at lower cost. Amtrak could not be expected to discern this opportunity with its present management. accounting and planning techniques.
These articles cannot cover these issues comprehensively. We can, however, explore specific case studies that illustrate Amtrak’s failure to address correctly cogent economic planning issues and how they would be addressed in a successful private business. We will illustrate how the Route Profitability System (RPS) contributes to counterproductive decisions in areas of equipment utilization, revenue generation, cost management and the 1983 station closure program.
The RPS System
Amtrak describes RPS as its only route accounting system. It accesses Amtrak’s cost data stored in their computer data bases, and is the source of virtually all of Amtrak’s purported short-term avoidable cost data used for planning and marketing. However, it does not and can not account for the actual costs of individual routes, trains or facilities, but instead calculates and projects average variable costs by collecting virtually all costs into pooled accounts, then allocating those costs back to various functions based largely upon management’s assumptions and preconceptions. These formulas must be simplistic due to inherent limitations in the computer data retrieval system and, by Amtrak’s admission, largely involve computation of averages and simple ratios (linear models) rather than use of even minimally sophisticated nonlinear computation techniques necessary for the correct marginal cost analyses discussed previously.
The RPS accounting system uses a computer language that is incapable of making the kinds of mathematical calculations that are indispensable to making accurate cost modeling calculations for financial projections.
We can apply simple marginal-cost, economy-of-scale and return-on-investment concepts to illustrate how badly flawed Amtrak’s accounting system is. The costs we will examine are fuel, food service, reservations, equipment utilization and scheduling and the station closure program. RPS’ failures in these easily measured areas are representative of its deficiencies throughout its 90-plus cost classifications.
The Folly of Averages
Average variable costs are intrinsically useless as a management tool for current operations or as a statement of the actual marginal cost of any proposed new train.
A taxi company in East Overshoe, N. Dak.. is owned and managed in Washington, D.C. It operates three identical cabs. Management is more absorbed by politics than managing the company, and fails to notice that their mechanic, Jake, who is paid $40,000 a year to maintain the cabs, has plenty of slack time and, worse, never touches cab No. 3, which only comes into the garage when Jake is taking his daily siesta. Jake’s supervisor spends all his time writing memos to Washington about his cost-cutting ideas and doesn’t notice, or covers up, Jake’s failure to service cab No. 3.
Management’s accountants have concluded the IRS will be satisfied if Jake’s salary and all parts and fuel costs are allocated to the cabs pro-rata based on annual car mileage. So. once a year, the supervisor orders Jake to read the odometers on all three cabs (Jake gets No. 3′s mileage by averaging the mileage from No.’s 1 and 2) and feeds the data into management’s government-approved accounting system. All costs are allocated to all three cabs even though No. 3 gets no maintenance and No.’s 1 and 2 get more than they need, and Jake could be reduced to part-time. Revenues are flat because cabs 1 and 2 are operating at capacity and No. 3 has increasing amounts of unreported down time as well as incredibly high fuel consumption caused by the deteriorating engine.
What will happen eventually? Cab No. 3′s revenues will drop off dramatically, and its real costs and down time will skyrocket, as it deteriorates from lack of maintenance. No accounting report will reveal the problem to Washington management. Jake and the supervisor keep getting paid. The government is content.
Is management’s accounting system adequate to allow it to detect and diagnose the problems in East Overshoe? Or to decide whether to retire or replace cab No. 3? Or to add a fourth or fifth cab? Or to fire Jake and/or the supervisor?
Fuel
Amtrak determines trains’ fuel costs not, as one would expect, by reading locomotive fuel meters or meters on fuel docks and multiplying by supplier prices, but by allocating its total fuel bill to trains in proportion to unit miles. To Amtrak, an F40 is an F40 is an F40, regardless of operating environment or consumption profile. Amtrak’s averaging of fuel bills probably results in overcharges to 403(b) trains, and masks any individual train’s actual consumption. Individual units’ fuel consumption rates can be an early symptom of costly mechanical problems. Amtrak’s system blinds them to this diagnostic opportunity.
Amtrak’s ignorance of actual individual fuel consumption leaves the corporation open to fuel waste and even theft. Everything is averaged, so discrepancies on individual routes cannot be identified by managers reviewing accounting reports, assuming the managers even knew enough about field operations to recognize a discrepancy if it appeared in a fuel consumption or cost report. Even if real fuel costs were monitored regularly at headquarters, it is likely significant overcharges or waste would not be discovered, because, incredibly, according to Amtrak’s senior financial officers, Amtrak’s financial system does not require that managers who prepare budgets bear responsibility for the financial results of operations as measured against those budgets. There is no internal check and balance to compare results to budget at a low enough level in the corporation to be meaningful. The implications of this failure transcend fuel management and infect every aspect of Amtrak’s financial management.
Lest anyone think this level of detail is unnecessary or unattainable for a railroad operating hundreds of diesel locomotives from numerous terminals, the nation’s major airlines have done exactly this, on their computers. for their aircraft fleets for years. Some know the precise fuel consumption of each engine on each aircraft over each flight segment on each day of the year. Among other things, monitoring fuel consumption patterns, engine by engine, provides early diagnosis of potential mechanical problems for turbines just as for diesels. Amtrak can and should achieve a modest proportion of that capability. Perhaps then they wouldn’t accidentally run out of fuel at Washington, D.C., on Christmas Eve.
Food Service
Amtrak determines the cost, hence the profit or loss, of its various food and beverage services by totaling food-service costs, then allocating those costs to various trains, regardless of the train’s actual costs, in proportion to the train’s food revenues! The actual food-service costs of each route, train or car are not known and not used. This is an astonishingly primitive, misleading and deceptive method of accounting for food service operations. It violates one of the fundamental precepts of an financial accounting systems, matching corresponding costs and revenues. It deprives management of a key tool to monitor and control the performance of individual operations. It makes meaningful cost projections impossible. It conceals the profits and losses of individual operations. It is easily manipulated. It conceals waste and theft.
Cars or routes with unusually high sales, hence higher productivity, are arbitrarily penalized by artificially being made to appear more costly. A particular route that benefits from an enthusiastic, effective, productive diner crew that stimulates more passengers to buy meals and to buy more expensive meals, cocktails and desserts will have disproportionately higher costs arbitrarily assigned to it precisely because of its higher revenues. For example, Minnesota’s North Star served 550 passengers in nine car trains in fall i 983 over two meal periods with a single lounge car with one attendant. A train like this sells a high volume of unusually high margin, thus very profitable, items such as liquor and snack foods, with relatively low labor and equipment costs, but will be artificially and falsely charged with larger costs precisely because of its higher revenues. Many high-revenue, long-distance trains catering to first-class and longer-distance riders (the average trip on Western trains spans three to five meal periods) will experience the same disproportionality as their passengers purchase more cocktails, snacks and higher-priced full meals.
Contrast a typical midday Metroliner, with only 50 or so passengers on board at any one time south of Philadelphia. It operates outside normal meal hours, but with the normal complement of two food-service cars in its 4-car train. The average rider is only on board for 90 minutes, and very little food will be sold. This Metroliner, nevertheless, would be assigned artificially low costs by RPS, despite the high actual costs associated with having two food-service cars, two attendants and no revenues to speak of, precisely because of its low revenue. The higher–cost, lower-revenue service would be made to appear far more profitable than it really was, or have its actual losses masked.
Even worse, on the now-defunct New England Metroliners, Amtrak gave away food free of charge. Because there is no food service revenue associated with it, the RPS formula used on national system trains could assign it no costs whatsoever, despite the obvious costs of the food itself and of the commissary and other employees who prepare and distribute it. The only redeeming virtue is that so few people rode (average ridership was less than 50 passengers total per trip) that relatively little food could be given away. The fare premium on these trains doesn’t even begin to cover the actual marginal cost of the free food and commissary labor (never mind train-specific advertising costs that alone exceeded $12 per passenger). In fact, Amtrak doesn’t know what these services cost. In December 1983, Graham Claytor wrote, “The extremely high premium fare charged on these trains has probably recovered almost all of the increased cost associated with the improved on-board service.” (Emphasis added.)
Graham Claytor justified stripping dinettes and lounges away from longer-distance Midwestern corridors to goldplate the NEC (where average trip lengths are less than half those in the Midwestern corridors) because of an alleged higher revenue per car-mile for dinettes operated in the NEC. Revenue per car-mile is one of the most useless statistics ever created. It tells nothing of profit, service, total sales volume, productivity, marginal traffic generating potential, return on investment, or any other meaningful index of performance. (Do Metroliners really need a dinette and club-lounge on each train to serve the 40 to 50 passengers on board at any one time south of Philadelphia? Did the New England Metroliners really need a full-service dinette?) By focusing on one meaningless index of performance, Amtrak is precluded from recognizing waste and opportunity.
If Amtrak’s accounting system were to be used in a food-service business in the private sector, the business would bankrupt itself in short order. Its high-revenue, high-profit restaurants would be closed because the accounting system made them appear to be too costly and unprofitable, particularly to a manager fixated on cost control without a proper appreciation of the basic concept that every opportunity has a correlative cost. Its high-cost, low-revenue, unprofitable units would be kept in service and expanded because the accounting system made them appear to be more successful by allocating their costs away to the high-revenue units. The business would fail with little warning when the firm’s capital ran out, and for reasons that would not be apparent from the firm’s financial statements. The failure would occur even at the hand of an unbiased manager. If the manager was prejudiced in favor of the high-cost, low-revenue units, and created or used the faulty accounting system to justify his favoritism, the failure would occur sooner and the total losses (and lost opportunities) would be even greater. This management pattern persists at Amtrak only because of the ongoing availability of needlessly large federal operating subsidies.
Amtrak’s food-service cost-accounting method was described recently to a financial executive of a major private-sector restaurant chain, who needs to know (and gets) the actual revenues and unit-specific costs of each restaurant in the chain to show which units are profitable, and which need special attention or should be closed. He reacted with total incredulity and then laughed. He did not believe that any multiunit food-service business could conceivably so manage its operations and remain in business, until the “business” was identified as Amtrak.
Reservations Service
RPS’ failure to state accurately the actual marginal cost of adding (or keeping) a train as “all reserved” in the Arrow system is graphically revealed by Amtrak’s abuse of the 403(b) mechanism with Minnesota’s North Star. This failure indicates the degree by which RPS’ average variable cost methodology overstates actual marginal costs generally. The magnitude of the overstatement of marginal cost here is a clear indication of why Amtrak refuses to inaugurate more new services in carefully selected new markets, where marginal operating profits could improve substantially the corporation’s overall revenue/cost ratio, achieve a positive rate of return on investment and reduce its need for federal operating subsidies, by means of expanded service.
Amtrak quoted Minnesota a price of about $447.000 as the state’s share of the cost of reservations service for two years of daily service. This figure represents, according to Amtrak, 65 percent of the short-term avoidable cost-i.e. what Amtrak claims is its actual marginal cost–of keeping the North Star as an all-reserved train. According to Amtrak. even that is a good deal for the state because Amtrak claims it only includes the direct costs of reservations agents and WATS lines in the STAC portion of Arrow costs, and that it treats computer costs and other reservation center costs (facility rent, maintenance, power, etc.) as fixed and not chargeable to individual trains. This means Amtrak thinks, or would have Minnesota believe, that the marginal cost of keeping the North Star as an all-reserved train for two years would be a whopping $688.000! This cost represents more than half of the total short term avoidable loss Amtrak projected for the North Star in this biennium.
Keeping the North Star all-reserved is critical because of the huge loads it attracts on weekends between May and October.
Reservations provide a necessary increase in the quality of service at what one would expect to be a lower cost than stationing protection cars in St. Paul every weekend. Stand-by cars represent millions of dollars of equipment that may or may not be used during Amtrak’s peak demand period. If the state declines to pay for reservation service, the train will revert to unreserved status. Amtrak then would turn away hundreds of customers every weekend (or at least on the first weekend), and incur the ill will of scores of standees.
We can test Amtrak’s RPS-based perception of the cost of this service. North Star ridership is essentially all round-trip excursion traffic, and heavily concentrated in summer and fall. Assume 100,000 annual riders translates to 50,000 round-trip passengers. These riders are predominantly excursion traffic, traveling as families, groups of friends, clubs, etc., so that most passengers travel in groups. Assume the average group is four people, and that only one member of each group telephones for the group’s reservation. Some will call twice, some will call only for information and some will deal with travel or station agents and not call at all. There may be no more than 50,000 calls per year to Amtrak for North Star reservations and information. Amtrak evidently believes, or expects Minnesota to believe, that it will incur out-of-pocket cash costs just for agents and telephone lines equal to roughly $13.75 per call to add the North Star to Arrow. This is, to say the least, an incredible overstatement of the actual marginal cost of adding the North Star to the existing reservation system. Even if these assumptions understate the number of telephone calls by 100 percent, Amtrak’s quoted price amounts to nearly $7 per call, which is still an incredibly high cost. The real average variable cost is probably closer to $3 per call: WATS lines cost about $20 per use-hour, and the fully allocated cost of a clerk should be about $15 per hour. producing about a $3 cost for a five-minute call. That cost approximates the cost per call of some airlines. But this still grossly overstates the marginal cost caused by adding the North Star to Arrow.
What would the actual marginal cost be? We know $688,000 is way too high. Amtrak’s figure implies that ten additional full-time agents and WATS lines are needed each year to handle just North Star call volume. Ridership this year will actually be closer to 85,000 on an annualized basis, so let’s analyze an annual call volume of 40,000 to determine the true marginal cost.
About 65 percent of North Star ridership is concentrated in five peak months from mid-May to mid-October. This would account for 26,000 calls. Allow five minutes per call, and spread them over 12 hours of each day of the five months, for an average hourly call volume of about 14 calls. Assume further that every Amtrak reservation agent (and telephone circuit) is already busy I 00 percent of his or her working day, and that not one call can possibly be handled by any existing staff (or lines) at any of the five regional reservations centers. Accordingly, each day two new agents, plus two part-time agents, would have to be employed or recalled and 11/2 new WATS lines installed, to handle North Star call volume over two shifts each day. For the other seven months, we can easily get by with two extra agents and one extra WATS line to handle an average call volume of 5 1/2 calls per hour. If the WATS lines cost $5,000 per month, and the agents $2,500 (including fringes and overhead), the total actual marginal cost of agents and telephone lines to handle North Star call volume is 22 agent-months and 14 1/2 WATS line-months per year, for a total first year cost of $127,500. If inflation, wage and cost increases in the second year come to 10 percent, the full two-year cost would be $267,250, or about $5.35 per call.
This illustration, however, still seriously overstates the true marginal cost. It assumed that not one North Star call could be handled by existing staff and telephone circuits, and that entirely new staff and line capacity was hired specifically because of the North Star’s call volume. Of course, in reality this would not occur. Calls would be bunched in the early evening hours, and existing agents and telephone circuits could handle a substantial percentage of calls for the train during the rest of the day. In reality, only during a handful of peak evening hours in the five peak months would limited part-time or overtime help be necessary to accommodate North Star calls over and above calls for other trains. The absolute maximum marginal cost caused by North Star calls, viewed as an increment to the existing system, would be a single, half-shift, peak hour (say 4-8 p.m.) reservation agent and maybe a single extra 100 hour-per-month WATS circuit, both limited to the May-October peak traffic period. The cost? About $3,800 per month (less than a dollar a call), less than $40,000 for two years, certainly not $688,000. Amtrak’s quote was 17 times greater than the likely true marginal cost.
So where did Amtrak get $688,000? If they didn’t just make it up, it may have come from dividing the total labor and telephone cost of the reservation centers by the time-weighted numbers of reservations and information calls received from the entire system and NBC in some prior accounting period. This figure would then be used to calculate an average variable cost of the Arrow system, which would be extended to the North Star in proportion to its total passenger count as a percentage of nationwide ridership. Amtrak admitted that its allocation formula for reservation costs for 403(b) billings was based on its assumptions, rather than on empirical data, much less real, causally related cost factors. One of those assumptions is that reservations costs increase in direct proportion to passenger boardings.
Amtrak refused to explain how it calculated the price quoted. Although the states are parties to contracts with Amtrak, Amtrak often refuses to disclose cost information claiming it is “proprietary information.” When confronted by MinnDOT on this issue (for the second time; Amtrak ignored MinnDOT’s first plea for relief for over two months), Amtrak “recalculated” the one-year cost for reservations service as just $5,000.
Against this background, Amtrak probably had good reason to refuse to explain the derivation of the charge. The National Conference of State Railroad Officials is on record as strongly concerned about Amtrak 403(b) billing and accounting. Did similar accounting distortions contribute to the demise of the Black Hawk, Spirit of California and other discontinued national system trains? In the case of Auto Train, Claytor said there would be no incremental costs for reservations because the system was already in place. If that is true, why is Minnesota charged anything at all?
Equipment utilization and scheduling
Amtrak’s failure to comprehend the complex interaction of costs, volume and revenues to maximize profit and return on investment has led to several examples of poor utilization of capital resources. Failure to accommodate Eagle standees with an available car due to concern about costs, as mentioned earlier, is one good example. Here are three more.
Amtrak had nine Superliners in dead storage at Beech Grove, some for over four years, while Western trains continued to turn away thousands of high-revenue customers six or more months out of every year. Another dozen Superliners were dedicated to North Star and Seattle-Portland equipment pools while all available Amfleet I cars were refitted for renovated Metroliner service, in a futile attempt to salvage low-revenue, end-point ridership lost to PeoplExpress and other discount air carriers. Superliner trainsets assigned to the North Star achieve a total use of just 306 revenue miles per day due to Amtrak’s extremely aggressive pricing of 403(b) services. The Reno Fun Train ties up another trainset for one Oakland-Reno round trip per week. Eagle and Sunset trainsets sit idle during one- and two-day turnarounds. You can’t make money with equipment standing idle.
This Superliner situation exists despite Amtrak’s highest revenue trains operating at capacity at least six months per year. It results from RPS’ prediction, which Amtrak management believes, that the cost to add a car to an existing train is essentially the same as the current average operating cost of a single car, rather than the actual marginal cost. Amtrak’s planning department believes that an add-on coach needs a 50 percent load factor ($6,000 in revenue on a single Chicago-Seattle trip, for example) to break even, even if no additional locomotives, food-service cars, or other major costs resulted. This is inherently implausible and, in fact, overstates the cost by at least 40 to 50 percent. To be extremely conservative, even by using Amtrak’s perception of costs, Amtrak loses over $20,000 per car per month of pure, genuine profit by shorting the consists of Western long-hauls by just one Superliner coach, just during the six peak travel months. Adding just one coach to each primary Western long-haul during these six months would earn as much as $10 million or more in profits per half year. Remember the widget factory’s lost profits discussed in the first installment.
In the Northeast, Graham Claytor has stated there is virtually no endpoint business on the Metroliners–traffic is predominantly to or from intermediate points. Ridership data appears to substantiate his claim, as do eyewitness accounts. So how does Amtrak structure operations in the NEC? Do they take advantage of this pattern to maximize revenues, equipment utilization and returns on investment? No. With the exception of a few Washington-Boston conventionals, the Corridor spine is structured for three endpoint niarkets-Washington-New York, Philadelphia-New York (primarily a glorified commuter service that produces nearly 50 percent of Amtrak’s NEC traffic) and Boston-New York. Through markets with longer trip lengths and higher revenues are not well served, and train sets are stopped and turned more often, with much lower utilization. Lucrative Northeastern markets readily available to Amtrak with no air competition, such as the inland route from Boston via Springfield, Mass., are ignored. Claytor wrote that “Many short-haul conventional services have relatively poor equipment utilization because of the relatively short trip times in comparison to terminal times as well as the peaked nature of demand.” Unfortunately, this statement is not true as a factual matter (not one of the short-haul budget or commuter airlines would tolerate or survive equipment utilization Amtrak accepts as inevitable) and it reflects a shockingly complacent management outlook. Turnaround times are completely controllable. The NEC could he served generously with probably half the equipment now used, freeing scores of cars for many other markets.
The New England Metroliners were structured to minimize ridership, equipment use, market share and revenues. Their structure made their only meaningful traffic source New York-Boston endpoint traffic, for which the four-hour, $45 train must compete with the one-hour air shuttle and $29 PeoplExpress jet services. Amtrak ignored potentially significant traffic opportunities for which there is no serious air competition by failing to operate these trains through New York to service such markets as Providence-Philadelphia or New Haven-Newark. They were scheduled so that even ambitious passengers could not connect at New York from anything, much less from another premium-fare Metroliner. As a result, the New England Metroliners carried 55 to 60 percent less traffic than the conventional trains they displaced, with ridership declining almost every month, despite a million dollar, train-specific advertising campaign. The premium fare isn’t an acceptable answer, because at obscene absolute fare levels it recovered less than 40 percent of the marginal cost of the service. This reflects Amtrak’s marketing philosophy in the NEC, which Graham Claytor said was “…improving profitability through higher yields on an existing traffic base rather than going for large volume increases.”
The schedules required two full trainsets (each representing at least $5 million worth of capital equipment) to sit idle for five hours during the middle of the business day at each terminal. You don’t make money with cars that aren’t running. An airline marketing officer who scheduled a DC-9 or 727 to be parked from noon to 5 p.m. each weekday would be fired. And Amtrak ended up having to park the New York turnaround train in Penn Station, taking up expensive track space, because they were unable to get the train back from Sunnyside yard, five hours later, in time for its 5 p.m. departure with reliable regularity. This leads to the same question about the NEC as the Superliners: How much is Amtrak losing on poorly utilized equipment? The Washington Metroliners do no better. Each $5,000,000 trainset ordinarily makes just one 440 mile round-trip each weekday. The new San Diego Metroliners repeat all these errors.
PART III
“[M]any expenses in running an Auto Train service that would be incremental to anybody else are not incremental to us. We can do ticketing and sales without any additional expense.” -W. Graham Claytor RAILWAY AGE, January 1983, Page 44
The station closure program
Graham Claytor coined a classic doublespeak term for this — “unmanning” stations. No matter what you call it, this program illustrates how Amtrak management ignores the concept of opportunity costs when making important decisions.
Early in 1983, Amtrak proposed to “cut costs” (to divert cash for the new train manager cost center) by closing revenue-generating stations serving “low-traffic” stops on the national system. Amtrak initially closed stations if station revenues failed a preset criterion. This ignored the numerous agents already being paid not to work under C-2, who could have operated these stations for several years without increasing Amtrak’s outlays or subsidy requirement. Revenue losses were not considered.
This criterion was fallacious. even apart from the effects on revenues and availability of agents under C2, for several reasons. Station revenue data was from a time when a mail-order ticket contest was conducted in the central reservation centers. Passengers calling the 800 number in some cases were told they could buy tickets only by mail, not at the stations-depriving them of revenue. At least two known instances involved tickets valued at several thousand dollars each, whose purchasers were quite surprised to find agents on duty when they arrived to board the train. The revenue criterion ignored travel agent sales, assuming station revenues alone accurately reflect station traffic and workload to check baggage, furnish on-time reports, adjust reservations, provide boarding assistance and so on. Onboard sales were also disregarded, even though the agent may have performed similar services. The criterion also ignored the sparse dispersal of sales representatives in the West, where agents act as de facto sales representatives. Needless to say, this criterion raised a din of protest.
Amtrak then announced a new criterion: the station’s “traffic level” measured by the number of coupons collected. Fifteen coupons per day would be the minimum to justify continued operation. This methodology is still fallacious and stupid. First, it discriminates against small stations without ticket printing machines. Each coupon printed by a machine (i.e., each segment on a trip) counts as a ticket, but multi-segment handwritten tickets count as just one ticket. It is unclear how group tickets would be counted. This policy indicates that management has no real comprehension of how their ticket system operates, or is simply incredibly inept at accurately determining patronage and its values.
This issue illustrates Amtrak’s failure to understand the basic business economic concepts reviewed in the first article. Amtrak consistently averages system-wide quantities and simply ratios them out linearly when making economic projections or comparisons, neglecting any effects of scale which result in volume selling at maximum profit. Management fails to recognize the actual cost of closing a station and, worse, disregards that closing a station may save x dollars of cost, but may forfeit 2x, 5x or more dollars in revenue. Conversely, manning an unmanned depot may add x dollars of cost, but could produce 2x, 5x or more dollars in revenues. Managers with a good understanding of the dynamic characteristics of the marketplace do not make this mistake.
Another problem was Amtrak’s primary reliance on head-count or coupon-count data, refusing to acknowledge the value of those heads or coupons. Crucial factors, such as passenger-miles, are not properly factored into the decision-making criterion. Using coupons as a lazy manager’s surrogate for passenger counts, only aggravates the failure. The fatal flaw is that Amtrak’s criterion ignores the differential revenues produced by passengers, fails to distinguish market segments and doesn’t consider connecting passengers. Incredibly, Amtrak’s ridership matrices do not account for connecting passengers. A Cleveland-to-Omaha passenger is counted and reported as two riders, one from Cleveland to Chicago and another Chicago-Omaha. This blinds Amtrak planners to the implications of cancelling one of the connecting trains, or adding new connections, and to the massive economy of scale potentially available in a national system network, Amtrak’s assumption that all coupons are worth the same average value is a predictable result of Amtrak’s lazy person’s aproach to management-dividing revenues by patronage. even within a single segment of a route, to determine average revenue per passenger, eventually inducing the perception that each passenger pays more or less the same.
In fact, most passengers outside the established corridors produce very high revenue per passenger. Many pay thousands of dollars to ride Amtrak and do not have to be pursued with million dollar ad campaigns. It is fallacious to equate those passengers to short-haul corridor riders to evaluate the viability of a manned station.
Amtrak later adopted another criterion: Station revenues had to be at least four times personnel costs to avoid unmanning. Even apart from the C-2 problem, this ignores the local costs of rent. utilities, insurance, ticketing, information and reservations. supervision and the effects of scale derivable from exceedingly modest investments in effective local advertising and sales promotion programs. It also indicates Amtrak still doesn’t know how to calculate each station’s own actual costs or potential revenues.
Consider this example. On the Eagle, a typical one-way fare is more than $80 per sale, so a station contributing “only fourteen” full-fare, one-way passengers per day would gross at least $1,200 per day or $408.800 per year. According to the ICC, only a third of that pays for direct train operations (fuel, crew, on-board services, track rent and all station costs). The rest goes to Amtrak’s tremendous overhead costs. At a typical single agent station, the agent earns about $30,000 for a characteristic single-shift-plus-overtime operation.
A non-NEC corridor station, where $15 fares are the norm, but with a “justifiable” level of 15 passengers per day, grosses about $225 per day or $81,000 per year. That barely covers the cost of two agents, and few, if any, corridor stations have as few as two agents. Triple the ridership and revenues are still less than half the “nonjustifiable” long haul station. Yet Amtrak’s criterion would close the $400.000 revenue source and leave the more costly $80,000 station open. Such a decision is nonsense at best, and at worst is evidence of persistent and intentional bias against national system long-haul services.
Annual revenues from a “justifiable” NEC station, where $6 fares, which about 40 percent of NEC riders pay, are the norm. could be as little as $32,850. The criterion would say that 15 passengers producing total revenues of $90 per day justify manning, when 14 long-haul passengers with revenues of $1,100 do not.
Consider the passenger counts necessary for corridor stations to achieve revenue or profit parity with the small-town long haul station. The non-NEC corridor station would have to produce over five times the patronage, or 80 passengers per day. to equal the revenue of 14 long-haul passengers. To achieve this head count. Amtrak would probably staff the station with three to six agents, driving the labor cost over $100,000 per year, $70,000 more than the single-agent station. Nearly 100 passengers per day at the corridor station approaches parity, without reference to advertising and other acquisition costs.
The small NEC station would have an even harder time, requiring between 180 and 225 passengers per day to equal 14 Western small-town long-haul passengers. If the small-town station had a modest increase in patronage, perhaps as little as one new rider every fourth trip. it would outstrip its corridor counterparts, particularly when the difference in investment to produce that increase is considered. But Amtrak either does not understand or has chosen to ignore this phenomenon.
This raises a significant question of resource allocation. Could Amtrak legitimately justify continuation of present staffing levels at large corridor stations based on corridor revenues alone? We seriously doubt it. There is simply too much disparity in the returns per passenger captured. Many NEC stations may be seriously overstaffed because they require no checked baggage service, could or do handle mostly passengers on multi-ride tickets, and could use automated ticketing just as their lowcost, short-haul airline competitors have done. Where is the justification for 80, 100, 150 or 200 on station staff when the vast majority of services do not require that kind of manpower? Over half the arrivals and departures at Los Angeles are this kind of corridor traffic, and stations like New York Penn Station and Philadelphia 30th Street handle an even higher percentage of these no-frills trains and riders.
Would Amtrak not have been better off laying off an incremental 10 to 20 employees, distributed over these large stations, and leaving the little towns, where the one agent is Amtrak’s sole presence and salesman, alone? Compare just the difference in revenues and costs and the answer will be apparent. True, corridor stations produce higher head counts (sometimes), but the small national system stations may get far greater revenues per dollar invested in employees and fixed facilities.
How many highly effective, informationoriented small-town weekly newspaper advertisements could Amtrak have bought for these “marginal” stations with the $13,000 they spent in one day in July 1983 for one unsuccessful New England Metroliner ad in the New YorK Times? How many stations could have been manned with the millions spent on the New England Metroliners9 I-low many train chiefs could have been hired with those millions with no station cutbacks out in the hinterlands or maybe even anywhere? Why can Amtrak management continually choose the wrong financial course and get away with it, to the detriment of most of the nation?
What would happen if Amtrak, for once, acted as a real business in the private sector would, using the basic principles of nonlinear economics, economies of scale, maximizing return on investment, and volume pricing at the margin?
An unmanned stop will have a baseline patronage, usually close to zero on a noncorridor route. If no labor or time-intensive changes on an existing train are associated with the stop, as is virtually always the case, and if there is available capacity on board (which Amtrak claims to be true despite frequent standing room only (SR()) conditions on long-haul trains), there are no additional train costs associated with changes made at the stop. Station costs thus can be considered independently here.
Amtrak’s systems would predict increased train operating costs from increased ridership, regardless of excess capacity. This is not accurate. Given excess capacity, even over just a segment of a route, to add train costs to the direct cost of the agent double-counts the costs because one-line stations’costs are already accounted in direct costs of train operations. Even adding cars, now idle, or transferred from other routes, does not produce a linear increase in cost.
If the unmanned station is changed by being lighted, heated and maintained by a caretaker, the better service will cause a small increase in patronage. The only difference from a properly staffed station is the lack of an agent, a telephone and minor cost items such as office supplies. Many of these stations don’t have or need a computer terminal. To be conservative, assume the agent costs $40,000 per year and the telephone and office supplies cost $1,500. At $80 per passenger, total station costs are recovered from a marginal increase of 519 passengers per year, or less than two additional full-fare one-way passengers per day over baseline! But that is extremely conservative because it still includes the caretaker. Thus two employees at train time, not one, are justified by this tiny increment of patronage.
Suppose the agent replaces the caretaker: breakeven now occurs at about 460 additional passengers per year rather than 519. This increase is just over one per day, or just one new round-trip passenger every other trip. Experience indicates the additional passenger is captured almost immediately from manning the station and is followed by many more. paying back the marginal cost of the agent many times over. This is the fallacy latent in requiring that revenues exceed an arbitrary percentage of station costs.
This approach will result in incredible paybacks even at stations with the lowest traffic levels, since any ticket sales beyond the one-per-trip breakeven point for the agent represents almost pure, genuine operating profit. If the agent deftly sells longhaul sleeper space, the return on the investment in the agent becomes phenomenal. The agent could pay his or her monthly salary with just one sale! The same analytical technique can be applied to computer terminals. advertising and other local improvements, to produce volume sales and maximum profit. because still greater paybacks will occur (up to the point of diminishing returns) with even moderate levels of traffic. This justifies further modest but highly profitable expansions in stations, station staff and even entire new routes in many markets. Amtrak’s methodology blinds management to this kind of profit opportunity.
Conclusions
What is at the root of these problems? We submit that Amtrak’s failure to understand how their own systems function coupled with use of meaningless and counterproductive planning and route-accounting methodologies, together with a singleminded fixation on cost containment, results in an attempt to starve the company into prosperity just as the farmer tried to do with his horse. This consistently fails in the private sector precisely because it ignores the correct relationship between costs, revenues and volume, and reflects an unrealistic and naive concept of corporate operations. While such oversimplified data analysis and business management are common in. the federal bureaucracy, they inevitably result in bankruptcy when applied to business corporations. Amtrak must be made to understand and implement correct business management techniques if the corporation is to develop any degree of prosperity, credibility and longevity.
Amtrak doesn’t have to be this way. It has the computer capacity and data systems to do the job right, if their systems were restructured and their executives knew what had to be done and did it.
When anyone wonders why Amtrak cuts or downgrades services or refuses to add services in the South, Midwest, sunbelt or West due to the alleged cost of the service, remember the 1700 percent overstatement of the true marginal cost of the North Star’s reservations service. When your accounting system overstates marginal costs of a proposed activity by a factor of as much as seventeen, it is little wonder management routinely refuses to authorize the proposed activity (remember the widget factory?). Add the overcharging of costs to long-haul food service, Amtrak’s misuse of equipment, cutting off its small town sources of big revenues, and the many other flaws of management outlook and route accounting, and it is a wonder any trains still run west of Harrisburg — or even east. Advocates are entitled to question how much revenue has been lost, how many trains discontinued and not inaugurated and how much federal operating support has been wasted through Amtrak’s accounting mistakes and manipulation. Advocates should demand this nonsense be ended, and that available resources be reinvested in highrevenue, high-return services that can earn operating profits to increase Amtrak’s revenue-cost ratio, reduce its operating deficit and political vulnerability, and rationally expand services throughout the nation.
