A bit of identification: I am a high school math teacher by trade, hold a master's degree in transportation science from the University of Washington, and have served as lead author on a series of books on Japanese rail transportation.

It is often taken for granted that U.S. urban rail transit systems, particularly those opened after 1945, provide few if any financial benefits to transit operators. Many authorities believe that the same or better performance could be obtained with non-rail alternatives. However, the 12 postwar rail systems considered below provide combined annual operating cost savings of about $800 million. There is no evidence for the huge net losses claimed by some critics.

The following analysis was developed from insights provided by E. L. Tennyson nearly ten years ago, and uses 1998 data from the Federal Transit Administration National Transit Database. It includes twelve U.S. urbanized areas where "new" rail systems have been opened since 1945. All transit operators in each region are considered in aggregate fashion. The regional average "total operating expense per passenger-mile" is calculated. A "no-rail" operating expense per passenger-mile figure is estimated, and is used to examine the annual financial impact of rail operations. "Rail" includes all light rail, heavy rail, and commuter-rail operations reporting to the Federal Transit Administration within each region.

The emphasis on operating cost per passenger-mile may be unusual in the U.S., but is not so elsewhere. The fundamental measure of the work performed by a transit system is the passenger-mile (or passenger-kilometer), not the "boarding" or the "unlinked trip." This fact is acknowledged and emphasized in various overseas publications. The fundamental unit cost of actually transporting people is total operating cost per passenger-mile. Moreover, it may be demonstrated that the "capacity" of a transit system is a function of passenger-miles, not "boardings."

(Skeptical readers are invited to consider the following exercise: Take the Seattle Monorail, which is almost exactly one mile long, and consider travel in one direction only. Set the maximum train frequency at one train per hour, and the maximum number of passengers per train at 100. The net "capacity" is therefore 100 passenger-miles and 100 "passengers" per hour.

(Now add a single intermediate station while maintaining all other parameters. Net "capacity" remains at 100 pass-miles and 100 "passengers" per hour.

(Now require that all passengers exit after riding between a single pair of stations (just as foreign passengers are required to do, according to some reports, on the subway system in Pyongyang, North Korea). Net "capacity" in terms of "boardings" suddenly doubles, to 200 pass/hr. However, net "capacity" in terms of passenger-miles does not change - and would not change no matter how many intermediate stations were added - without increases in permitted service frequency or vehicle capacity.)

Failure to address the average distance traveled by current and prospective transit passengers is a chronic omission in U.S. urban transit planning and analysis. Distances between origins and destinations are inherent in the data used for travel-demand forecasting. However, this information seems to "get lost" during the process, leading to unrealistic results. For example, Seattle's current light-rail forecast implies an average travel distance (per boarding) well below that carried by the current bus system - or any other "postwar" light-rail facility in the U.S. If more-realistic average travel distances are used in a simple "reality-check," the forecast plummets.

The problem is not confined to pre-construction planning. A recent analysis of Atlanta's rail system purports to find that much greater patronage could have been obtained at much lower cost with an expanded bus system. However, the patronage estimates for the no-rail "alternatives" were completely independent of average travel distance. In other words, passengers who were forecast to board would not necessarily travel.

The author does not intend to imply that statistics based on passenger counts ("boardings") have no utility, for this is not the case. However, he does assert that inaccurate results are likely to occur when the issues related to transit passenger travel distance are ignored.

The analysis presented below has a significant problem related to credibility: very large numbers which favor rail transit may be generated using a simple (if perhaps unfamiliar) technique. Perhaps the best illustration uses San Francisco's BART as an example.

BART's annual traffic load is approximately one billion passenger-miles. In 1998, the system reported an annual total operating cost of $0.30 per passenger-mile. There is very little evidence that a "no-rail" alternative transit system (a regional express-bus network, perhaps) could achieve the same results for less than about $0.40 per passenger-mile. This exercise assumes that the same results (same number of passengers per year, each traveling the same distance) could be achieved - somehow - without the BART rail network.

The unit operating-cost difference between "rail" and "no-rail" alternatives is one thin dime -- $0.10 per passenger-mile. However, this implies an annual operating-cost saving of $100 million. This implies in turn that the system provides large savings to Bay Area taxpayers and transit riders - and has paid for itself through operating-cost savings alone.

Americans seem to distrust conclusions based on passenger-miles - perhaps the phrase "cost per passenger" sounds more democratic. In addition, certain critics of public transit and rail transit tend to denounce anything that demonstrates that rail transit might be a good idea. However, if transit service could be provided on a strict "demand" basis (no more seats than filled, and no more miles than utilized), it would be clear that greater costs accrue to transit operators if passengers travel over longer distances. Once again, cost per passenger-mile is the fundamental unit cost of actually moving people.

I. OPERATING COST SAVINGS FROM POSTWAR U.S. URBAN RAIL SYSTEMS

The attached table combines all transit operators in a region, determining total operating expense, boardings and passenger-miles for 1988. "Total Operating Expense per Passenger-Mile" is a regional average.

The exercise of determining operating-cost "savings" as the result of rail transit is, to a large extent, purely academic. There is no simple way to estimate the number of passengers that would be "lost to transit" (as the "Light Rail Central" webmaster put it) in the absence of even the most lackluster rail systems. Lower ridership would result in lower total cost, of course, but this would have to be weighted - somehow - against the impact of higher auto traffic and so forth. Considering the 12 metropolitan regions one by one:

--Atlanta's remarkably efficient rail system provides large savings over an all-bus system, although estimating what these might be is difficult. If a "no-rail" system could 1.) hold on to MARTA's current traffic, and 2.) match the performance of Houston's METRO, $0.35 per passenger-mile (both, in the author's view, are unlikely), the implied annual net "saving" is about $15 million. At $0.40 per pass-mi, an all-bus system would cost roughly $53 million more per year to operate. At $0.50 per pass-mi (which the author believes is closest to what an all-bus system in Atlanta might achieve), annual net saving would be roughly $130 million. This is more than 50 percent of MARTA's current annual (1998) operating budget.

It pays to note that less than 40 percent of MARTA patrons are "captive" riders, compared to more than 70 percent in Houston. A large majority of Houston transit passengers are from low-income households, compared to roughly 40 percent in Atlanta.

--Baltimore is one city where rail does not appear to provide much advantage(although all MARC commuter traffic was assigned to Washington for lack of any way to separate Washington and Baltimore traffic and costs). MTA posts an annual average of $0.53 per passenger-mile. The implied annual $4 million saving is just 2 percent of MTA's annual budget. However, one must ask how much of MTA's current traffic would move by some other mode if the Metro Subway and Central Light Rail lines did not exist - and how much more might be carried if the urban rail network were more extensive.

-Buffalo's rail line provides no apparent operating-cost saving. In fact, this analysis implies that NFTA's annual costs are 2.5 percent higher than they would be without the rail line. However, it is possible to demonstrate that the performance of the rail line would be greatly improved once completed to Amherst (see below). And, once again, we do not know what percent of NFTA's current ridership would be "lost" without the rail line.

--In Dallas, the DART rail network costs $0.40 per pass-mi less to operate than its buses. Given an overall "no-rail" cost of $0.80 per pass-mi ($0.10 more than the current mixed-mode network), implied annual cost savings is nearly $31 million. This is 14 percent of the total regional transit operating budget.

--Rail does not change overall transit operating costs in the Miami - Fort Lauderdale - West Palm Beach region. However, the "regional" approach masks the fact that Tri-Rail "savings" offset higher costs of the Metrorail /Metromover network. Once again, one must ask what share of current ridership would be "lost" without rail, and how the impact of this "loss" should be evaluated.

--How would transit operating costs in the Los Angeles region change without rail? At 1998, it would be difficult to make a case for an overall "no-rail" cost per pass-mi more than $0.05 greater than posted by the (then-existing) network. But this, owing to the immensity of the annual transit traffic load(2.5 billion passenger-miles), implies more than $100 million in annual cost savings. (Even a $0.03-cent difference implies more than $75 million in annual cost savings.) Completion of the Red Line to North Hollywood has probably pushed the "rail" share of overall annual pass-mi to roughly 25 percent (together with a significant reduction in operating cost per pass-mi as the line has finally attracted the traffic load for which it was built). Unfortunately, we will have to wait until the National Transit Database statistics for 2001 are out until we can assess the impact of the extended Red Line on L.A. transit operating costs.

--Portland's $0.03 per pass-mi cost difference between Tri-Met rail and bus operating implies a $10 million annual cost saving, or 6 percent of the 1998 operating budget. The impact of the Westside cannot be assessed until 1999 statistics are available. Cumulative cost savings for Portland are estimated below.

--Results are similar in Sacramento - an average "no-rail" cost of $0.50 per passenger-mile works out to a $5 million annual cost saving, or 8 percent of the total operating budget.

--San Diego's implied annual cost saving, given an overall "no-rail" operating cost of $0.40 per pass-mi, is $42 million, or 29 percent of the regional transit operating budget.

--BART, San Francisco Muni, San Jose VTA and the Caltrain commuter-rail line are aggregated in the San Francisco Bay Area. The implied annual cost saving resulting from rail operation - given an overall "non-rail" cost of $0.60 per pass-mi - is roughly $300 million per year. This is 28 percent of the total annual operating budgets for the region's transit systems.

--Cost savings for St. Louis are estimated in detail below. This exercisesubstantially reinforces a "no-rail" operating cost of $0.60 per pass-mi. This implies in turn an annual cost saving of 30 million, or 25 percent of the regional transit operating budget. Note how "that one line" carries a staggering 39 percent of the region's entire transit workload.

--For Washington, DC, a "no-rail" cost of $0.60 per pass-mi implies an annual total cost saving of $274 million - fully 36 percent of the region's overall annual transit operating budget. Here is a prime example of a region where one must ask what share of today's total transit traffic could a "no-rail" network manage to carry under the best of circumstances.

In sum: It is true that comparison of operating cost per pass-mi between operators is complicated by various underlying factors. Perhaps the most important of these are 1.) relative labor costs, and 2.) "service effectiveness," the average passenger load carried per vehicle-mile.

However, it would appear that postwar rail systems in the 12 regions above provide annual operating cost savings on the order of $800 million. There is no evidence for the huge net losses claimed by some rail critics. If anything, this analysis is conservative, as it considers only direct benefits to transit operators in terms of cost savings. If direct and indirect benefits to consumers and society were considered (and could be estimated with reasonable precision, the figure above would be much higher.

CITY Op Pass Pass Tot Op Exp Rail %age Exp (mill) -Mi per Pass-Mi of Total (mill) (mill) Pass-Miles

Atlanta $255 161 762 $0.33 64% Baltimore 207 101 401 0.52 29% Buffalo 64 27 80 0.80 20% Dallas-Ft. W 212 65 307 0.70 21% Miami-W Plm B 298 112 593 0.50 29% Los Angeles 1,061 580 2,522 0.42 19% Portland 161 86 338 0.48 19% Sacramento 60 29 129 0.46 31% San Diego 146 93 472 0.31 38% S.F. - San Jose 1,123 472 2,398 0.47 53% St. Louis 118 54 246 0.48 39% Washington, DC 759 371 1,715 0.44 70%

SELECTED ALL-BUS REGIONS

Austin $52 30 97 $0.54 ---- Cincinnati 64 32 168 0.38 ---- Columbus 49 18 74 0.66 ---- Detroit 195 66 269 0.72 ---- Honolulu 99 72 359 0.28 ---- Houston * 185 94 522 0.35 ---- Indianapolis 20 10 47 0.43 ---- Kansas City 38 15 57 0.67 ---- Louisville 36 17 59 0.60 ---- Milwaukee 101 72 195 0.52 ---- Mpls - St. P * 151 66 251 0.60 ---- Phoenix * 64 35 139 0.46 ---- Slt Lke Cty * 62 24 101 0.61 ---- San Antonio 64 41 165 0.39 ---- Seattle * 339 112 621 0.54 ---- Tampa - St. P 48 18 79 0.61 ----

Seattle statistics include King County Metro, Community Transit, Everett Transit, and Pierce Transit. * Rail system authorized, or open after 1998.

II. OPERATING COST SAVINGS: FOCUS ON ST. LOUIS

In 1993, its last "all-bus" year, the Bi-State Transit Agency recorded 40 million boardings and carried 173 million passenger-miles. Its (inflation-adjusted) operating expense was $94 million (in 1998 dollars).

In 1998, the most recent year for which statistics are available, Bi-State recorded 54 million boardings, up 35 percent over 5 years. Passenger-miles were up by 42 percent, to 246 million. Operating costs were up as well, but only by 25 percent (to $118 million; in 1998$s). Inflation-adjusted total operating expense per boarding fell by seven percent (from $2.35 to $2.18), and total operating expense per passenger-mile fell by 11 percent (from $0.54 to $0.48).

According to statistics compiled by the American Public Transit Association, inflation-adjusted (nationwide) average operating expense per bus passenger-mile rose by 8 percent between 1988 and 1996 (to $0.54). But average operating expense per heavy-rail passenger-mile fell during the same period -- by 21 percent (to $0.30). Average operating expense per light-rail passenger-mile also fell, by 10 percent (to $0.45).

Between 1992 and 1996, in urbanized areas having more than one million residents, average (inflation-adjusted) operating expense per bus passenger-mile rose by four percent (to $0.54). Meanwhile, rail costs fell, by 17 percent for heavy rail (to $0.30 per pass-mi), and by six percent for light rail (to $0.45 per pass-mi).

The increase in passenger traffic carried by Bi-State, accompanied by the decrease in unit operating costs actually recorded, could not have been obtained without the light-rail facility.

Adjusting for inflation, the estimated annual operating-cost saving of 30 million represents an annual rate of return on light-rail construction cost of about 4 percent. This implies that the system would pay for itself in 25 years out of operating-cost savings alone. The true "rate of return" would have to include direct and indirect benefits to consumers and society as a whole. Estimating the dollar value of these benefits with reasonable precision would be a daunting task. (A direct financial return of seven percent was considered the "benchmark" back in streetcar days.)

If this estimate is valid, then one could make an argument for increased investment in light rail as follows:

1. To the extent that economists agree on anything, they tend to state the "real" historic long-term cost of capital (separate from a.) inflation, b.) risk premiums and c.) inappropriate assessment of "opportunity cost") at about three percent.

2. Although there is no agreement on how to quantify them, travel-time savings to users, providing an attractive alternative to auto use, increases in transit's market share, improvements in regional quality of life and other light-rail benefits certainly have value - positive dollar value.

3. Therefore, a light-rail system able to return at least the long-term cost of capital from operating cost savings alone should be worth building.

III. CUMULATIVE OPERATING COST SAVINGS IN PORTLAND:

Portland's light-rail system does not rank among the "high-cost" systems, for it cost significantly more per mile to build than the San Diego South Line or the existing Sacramento system. Nor is it the most efficient postwar light-rail system, owing to the two-car train length limit and a vehicle fleet which was not adequate for peak-hour traffic levels for the first 12 years of operation (below).

However, the Tri-Met system has achieved remarkable results. Between 1990 and 1997, the regional population increased by 18 percent and auto use ("vehicle-miles of travel") increased by 24 percent. Tri-Met boardings, however, increased by 30 percent. Portland is the only U.S. metropolitan region where transit ridership is growing faster than auto use.

In 1974, transit ridership in Portland was near its low point. Ridership fell from 60 million revenue passengers in 1950 to 18 million in 1969. The 1974 revenue passenger count was just below 21 million. Adding "transfer" passengers for compatibility with today's statistics gives 25.8 million annual boardings in 1974. Total operating expense, adjusting for inflation, was $50.2 million (in 1999 dollars). Operating cost per boarding was $1.94 (1999$s). Operating cost per passenger-mile was about $0.45.

In 1999, the annual boarding count was slightly more than 81 million - up by nearly 215 percent over 26 years. But despite large increases in labor and fuel costs since 1974, the annual operating cost per boarding is actually six percent lower -- $1.82. The annual operating cost per passenger-mile is unchanged ($0.46). That, given the spiraling increases in transit operating costs over the past 25 years, is truly a remarkable performance.

(Average cost per boarding is lower because the average travel distance per boarding is slightly less than in 1974. This is a mixed blessing, as discussed below.)

Tri-Met operating cost per annual vehicle-mile did rise from $3.47 (again in 1999$s) to $6.24 between 1974 and 1998 (the 1998 figure for buses alone was $5.69). However, the system maintained its overall cost efficiency by increasing the effectiveness of its service. In 1974, the system carried 7.7 passenger-miles per revenue vehicle-mile. This awkward phrase simply describes the average number of passengers carried aboard each vehicle while in service. The number is lower than one might expect since it includes all service year-round: peak, off-peak, late nights and weekends, on trunk and feeder lines combined. But in 1998, the system managed 12.9 pass-mi per veh-mi overall. The figure for buses alone was 11.0; that for rail alone was 37. Obviously, the rail system does not account for all of Tri-Met's cost efficiency.

Tri-Met's ridership growth was slowed by the oil-price collapse of the mid-1980s and the depressed state of the regional economy. In fact, overall ridership began falling shortly before the Eastside MAX light-rail line opened. In 1986, the system's last "all-bus" year, annual boardings reached 50.3 million. The operating budget was 94.3 million (1999$s); cost per boarding was $1.87 and per passenger-mile was $0.54.

Ridership systemwide remained static, or nearly so, for the next few years. But the rail line opening brought a sharp drop in unit operating costs: $1.54 per boarding, and $0.42 per passenger-mile.

Rail ridership grew each year, except between 1992 and 1993 when it remained static. Bus ridership began an upward trend in 1990, but slipped back slightly from 1992 to 1993, from 1996 to 1997, and from 1998 to 1999. The net result: an overall upward trend from 1990, broken by periods of "leveling off."

Tri-Met's unit cost statistics fluctuated up and down during this period. There was a clear upward trend in "input" costs: inflation-adjusted cost per vehicle-mile rose from $4.96 in 1990 ($4.97 for buses alone) to $6.24 in 1998 ($5.69 for buses alone). However, there was no similar trend for "output" costs: per boarding or per passenger-mile. This reflects the success of the Tri-Met management in maintaining the system's cost-effectiveness. It avoided the upward trend in bus operating costs per pass-mile: four percent between 1992 and 1996 (to $0.54; 1996$s), for urbanized areas having more than one million residents (according to APTA statistics). Tri-Met managed to hold its bus operating costs to $0.50 (1996$s) per pass-mile during this period. Unfortunately, light-rail costs rose from $0.30 to 0.38 per pass-mile (still below the national average of 0.45 per pass-mi).

Cumulative operating cost savings as the result of rail operation (based on a "null" or "do-nothing" alternative) was estimated based on the operating cost of an "all-bus" system for each year between 1987 and 1998. This was done by using the actual bus pass-mile, rail pass-mile and bus operating cost per pass-mile for each year. The estimated-savings figure is $110 million in 1999 dollars, or slightly more than $9 million per year. This implies a three percent rate of return on the cost of building the system(adjusting for inflation).

The savings might have been greater if Eastside line ridership growth had not been shackled by a too-small car fleet. After opening, off-peak ridership almost immediately required two-car trains through most of the day. This was not anticipated prior to opening. It also became clear that the fleet was not large enough to accommodate all potential peak ridership at a crowding level which Portland transit riders would tolerate. But no additional cars were ordered until the Westside line was built. Eastside service was substantially increased concurrent with the Westside opening. Vehicle-miles were increased by 31 percent (and train-miles by 19 percent). One month later, peak-hour ridership had increased by more than 17 percent. Overall weekday ridership was up by 34 percent over levels recorded two years previously. Some of this traffic was obviously generated by the new Westside line, but it is clear that a larger initial fleet would have led to greater ridership during the first 12 years of light-rail operation in Portland.

Annual cost savings provided by light-rail operation jumped from $10.2 million in 1998 to approximately $30 million in 1999 (1999$s). Despite the relatively high cost of the Westside line, this still implies an annual rate of return of about three percent, adjusting for inflation. Thus, both lines will pay for themselves within the standard 40-year life cycle from operating-cost savings alone.

The average distance traveled by Portland's transit patrons is static, and has remained so over 25 years. This suggests that the bulk of the ridership growth has been generated by relatively short-distance passengers. Work trips by auto are, in general, longer than work trips by transit, and Tri-Met may have experienced a relative lack of success in attracting passengers from autos. On the other hand, lack of growth in average travel distance by transit may also reflect the region's success in coordinating land-use planning with transit, controlling sprawl and confining new development within its "urban growth boundary." Further research is indicated.

IV. OPERATING EFFICIENCY OF RAIL TRANSIT AS TRANSIT TRAFFIC RISES

In St. Louis, Bi-State recorded 8 million rail boardings in 1994, the first full year of operation. The line carried 43 million passenger-miles, and cost $12.3 million (1998$s) to opeate. In 1998, rail boardings were up by 82 percent, to nearly 15 million. The passenger-mile count soared to 96 million, a 125 percent increase. Meanwhile, annual operating expense grew by 56 percent, to $19 million. As a result, total operating cost per passenger-mile fell by 31 percent, from $0.29 to $0.20 (1998$s).

In Los Angeles, the Metrolink commuter-rail network increased its annual boarding count by 14 percent (to 6.2 million) between 1996 and 1998. Average weekday ridership was up by 11 percent (to 23,600), while annual passenger-miles increased by a whopping 41 percent (to 272 million). The system obviously succeeded in capturing more, and longer, trips. Meanwhile, inflation-adjusted operating expense rose by only 3.4 percent (to $56.8 million; 1998$s).

Operating expense per vehicle-mile fell by 20 percent (to $10.00), while operating expense per vehicle-hour fell by 18 percent (to $411.15). Commuter-rail operating expense per vehicle-hour may appear fearsome, but are spread over a large number of miles (owing to high operating speed) and passengers (owing to large vehicle size). Total operating expense per boarding fell by 9.5 percent (to $9.14), while that per passenger-mile - the fundamental measure of a transit system's performance - fell by 25 percent(to $0.21).

Clearly, the operating efficiency of rail transit increases as traffic rises. The opposite tends to be true of unimproved all-bus systems.

V. POTENTIAL PRODUCTIVITY INCREASE THROUGH RAIL EXPANSION IN BUFFALO

Completion of Buffalo's current rail line from South Campus to Amherst would attract new ridership and increase the operating efficiency of the rail line together with the overall NFTA transit system.

The 6-mile Amherst segment would attract approximately 20,000 additional weekday boardings (with roughly 14,000 accounted for by State University of New York at Buffalo students). Much of the new travel would occur between the two SUNYAB campuses. However, other travel should raise the average travel distance per boarding from 2.3 miles (1998) to at least three miles.

Using the "least favorable" operating-cost assumptions (no change in total operating cost per vehicle-mile; only a modest increase in annual vehicle-miles per route-mile), annual operating cost per rail pass-mi would fall by nearly 20 percent. Annual operating cost per pass-mi systemwide would fall by nearly 5 percent, producing a potential $5 million in annual operating-cost savings over an unimproved all-bus system.

A three-mile average travel distance for a 12-mile radial suburb-to-downtown rail line in the U.S. is implausibly low. If similar calculations are performed assuming an ATD of four miles, the total operating cost per rail pass-mi falls by nearly 40 percent, and that per systemwide pass-mi falls by nearly 20 percent. This implies an annual operating-cost saving (over a "null" alternative) of nearly $20 million per year. This amount is roughly 4-5 percent of the projected construction cost of completing the line to Amherst.

NOTE: The example used below to emphasize the "credibility gap" is Bay Area Rapid Transit, which carries more than a billion passenger-miles per year by now. Its current total operating cost works out to about $0.30 per passenger-mile. There is absolutely no evidence that a "BART-substitute" express bus system could handle this traffic load for less than $0.40 per passenger-mile. That's assuming, of course, that any non-rail public transit system could carry BART's traffic volume. Multiply the unit operating cost difference by the annual traffic volume, and the cost saving works out to $100 million. That's $100 million per year.

The method I used below is very simple: I compared overall regional transit operating cost per passenger-mile with a hypothetical "non-rail" figure in order to estimate an annual cost saving. For two cities, St. Louis and Portland, I was able to estimate cumulative cost savings as the result of rail transit. Results, as you'll see, are mixed, but generally favorable.

in Buffalo, NY
thru
the Expansion of Mertro Rail