Detroit’s "SpeedLink"
Bus Rapid Transit (BRT) Proposal:
A Critique of Operating Cost Estimates Working Paper 02-01
Leroy W. Demery, Jr. • Contributions by Michael D. Setty • 2002 • Updated Sept. 22, 2007
Copyright 2005-2007,
Detroit’s Metropolitan Affairs Coalition (MAC) outlined a conceptual network of six radial and five peripheral bus rapid transit corridors in a report published in 2001 (“SpeedLink: A Rapid Transit Option for Greater Detroit," (2007.9.22), large file, 2.3mb).
Based on this analysis, it appears that MAC has greatly underestimated likely SpeedLink operating costs. Likely operating costs for the proposed services (based on "exclusive lane arterial operation" and "operation of low-floor articulated vehicles") are twice to three times the amounts estimated, because of:
1.)    Operation of a 100-percent articulated fleet.
2.)    Increased energy consumption due to increased vehicle-miles per service hour.
3.)    Reduced energy efficiency due to higher operating speed.
4.)    Underestimation of service levels required for the projected ridership.
5.)    Omission of station operating, maintenance, fare collection and security costs.
The cost-effectiveness of SpeedLink with respect to other alternatives disappears given sufficiently large differences in operating cost, as detailed below (Sections 5 and 6).
Credibility is the critical issue for supporters of major transit investments, as demonstrated repeatedly in various U.S. cities over the past 25 years. The SpeedLink proposal is not likely to secure, and retain, the necessary foundation of public and political support if questions regarding operating costs are ignored or papered over.
Operating cost per revenue service hour (RSH) is the most useful performance measure for transit service planning. MAC estimated SpeedLink operating costs assuming $100 per RSH. This assumption is supported by results obtained by the two transit operations serving Metro Detroit: City of Detroit, Department of Transportation (DDOT), and Suburban Mobility Authority for Regional Transportation (SMART). At 2000, DDOT posted unit operating costs of $99.90 per RSH; SMART posted $102.02 (National Transit Database). Therefore, $100 per RSH, which "represents the high end for BRT" according to MAC, appears a reasonable "baseline" for SpeedLink. However, MAC did not consider the factors outlined above. As documented below (Sections 1-3), SpeedLink would cost about $140 per RSH to operate. This figure is exclusive of station lighting, maintenance, and cleaning costs, security costs, and costs associated with self-service fare collection. These "infrastructure" or "overhead" costs would raise the overall SpeedLink operating cost up to the range of $170-190 per RSH.
It is essential to understand the difference between "average" and "marginal" (or "incremental") operating cost. Marginal cost refers to the expense accruing to an existing transit operator when it adds service. Marginal cost does not include overhead, and so does not reflect the total cost of providing service. Marginal cost estimates are of little use when planning a new service that will not be operated by existing operators or by existing modes. Such estimates are also likely to understate the cost of large-scale service expansion by an existing operator.
A recent U.S. General Accounting Office report ("Mass Transit: Bus Rapid Transit Shows Promise," , 2007.9.22) compared marginal operating costs for BRT with average (i.e. total) operating costs for light rail (LRT). This is a serious conceptual error that others would do well to avoid. MAC recognizes the problem and avoids the error: "SpeedLink will not likely be an incremental increase to an existing bus operation but a new bus operation with minimal economies of scale."
1.) Additional Cost Per RSH: Articulated Buses
Assumption of the same cost per RSH for standard and articulated buses cannot be justified. Articulated buses cost 25 percent more to operate than standard buses, owing to higher fuel consumption and other factors (Figure 1-1, below).
Note: Various operators reported that articulated bus maintenance costs were 33-50 percent greater than for standard buses, except Seattle, which reported that maintenance costs were equal, or slightly lower.
Source: "Articulated Buses A Planning Handbook," UMTA 1984.
Articulated buses have a lower power-to-weight ratio than standard buses and consume 10 percent more fuel ("Articulated Bus Report," UMTA 1982).
It is true that articulated buses achieve greater unit productivity owing to larger vehicle size, but that is not relevant to the matter at hand, which is unit operating cost. Considering the higher operating cost of articulated buses per RSH, the minimum defensible operating cost estimate for SpeedLink is about $125 per RSH.
2.) Additional Cost Per RSH: Additional Veh-Mi Per RSH
A typical transit planning assumption holds that operating cost per RSH is not influenced by the operating speed. This is not true, although it may reflect actual experience given the typical U.S. transit service pattern:
a.) frequent stops and low "cruise speed," the average operating speed between stops, or
b.) infrequent stops and high cruise speed.
However, given a transit service providing high cruise speeds between frequent stops, this rule of thumb will lead to substantial underestimation of operating cost.
MAC has designated Woodward Avenue as the SpeedLink "starter" corridor. This would extend 25.3 miles between downtown Detroit and downtown Pontiac, with 31 stations. The "average operating speed" ("annual service miles" per RSH) assumed for operating-cost estimation is 23.2. This, adjusted for the 15 percent recovery time assumed by MAC, implies a "schedule" (passenger) speed of 30 mph. The implied cruise speed is in the 35-45 mph range, given the average spacing between stations of 0.8 mile.
It is true that fuel expense accounts for a small fraction of bus operating cost per RSH. However, given an increase in operating speed, fuel would account for a relatively higher percentage of expense per RSH even if fuel consumption per mile did not change. If, for example, operating speed were doubled (implying twice as many vehicle-miles per RSH), fuel would account for 9.5 percent, rather than five percent, of total operating expense (Figure 2-1, below). The total cost per RSH also increases, so the "share" of the "increased" cost represented by fuel expense does not quite double.
Note that the example above assumes that fuel consumption per mile remains constant as speed is doubled. Even if fuel economy (miles per gallon) improves with speed, per-hour fuel consumption would still increase. In order to maintain a fixed rate of fuel consumption per hour given a doubling of operating speed, the vehicle would have to consume 50 percent less fuel per mile.
It is striking that planning software incorporates the erroneous assumption that fuel expense is not influenced by operating speed. For example, see "FuelCost 1.0"
 ( 2007.9.22 ), designed specifically for transit buses, has only one model input related to fuel economy: "miles per gallon." The default is set at 4.0 mpg, based on "typical" experience with diesel buses in transit service. This factor may be changed by the user to reflect an "easier or tougher duty cycle." However, FuelCost has no internal checks that would compare, for example, vehicle-miles to vehicle-hours. FuelCost users might underestimate fuel consumption by a large margin by using the "default," without considering that fuel economy changes with operating speed and the number of acceleration cycles per mile.
DDOT achieved (at 2000) 11.4 revenue vehicle-miles per RSH, and SMART achieved 17.8. Allowing for recovery time, these figures suggest passenger speeds ranging between 15 and 25 mph. This in turn suggests cruise speeds in the 25-35 mph range. The cruise speed range implied by the SpeedLink operating-cost estimate is 40-80 percent higher.
Considering 1.) the higher operating cost of articulated buses per RSH and their greater fuel consumption, and 2.) the higher fuel consumption per RSH associated with an increase in revenue vehicle-miles per RSH, the minimum defensible operating cost estimate for SpeedLink is about $130 per RSH.
3.) Additional Cost Per RSH: Cruise Speed and Energy Consumption
The energy-consumption figures in this section stem from a computer simulation conducted by Sims and Miller (1982), whose important findings have received little attention. The examples illustrated below (Figure 3-1) were based on results presented by Sims and Miller, and assume two stops per mile:
25-mph cruise speed: fuel expense accounts for 5 percent of operating cost per RSH (typical planning assumption).
35-mph cruise speed: fuel expense up by 18 percent per mile (from 25 mph).
45-mph cruise speed: fuel expense up by 55 percent per mile (from 25 mph).
LRT energy consumption remains relatively stable over a wide range of stop frequencies and cruise speeds. A "local" transit service with 13 stops per mile and a 25-mph cruise speed is far more demanding than an "express" service with just under two stops per mile. But LRT energy consumption varies little: 9 mpg equivalents for the "local," and 10 mpg equivalents for the "express" scenario. In other words, the "local" service consumes 10 percent more power per mile than the "express" service.
In contrast, bus energy consumption for the "local" scenario above is 2 mpg, and that for the "express" scenario is 3 mpg. That is, the "local" service consumes 50 percent more fuel per mile.
Consequent increases in overall operating cost per RSH depend, of course, on the number of revenue vehicle-miles per RSH. Such increases are not offset by reduced platform labor expense. Unlike operating cost per vehicle-mile, platform labor cost per RSH does not vary with speed. "System" operating costs may be lower if the operator manages to reduce the number of vehicles in maximum service (i.e. the number of vehicles required for peak period service).
Notes: Relative fuel consumption based on two stops per mile.
Source: Sims and Miller (1982).
On the other hand, if stops are few and far between, one can provide passenger speeds in the 40-50 mph range and achieve remarkable energy efficiency. A few long-distance express-bus operators in Greater New York accomplish this. Considering all three factors elaborated above, the minimum defensible operating cost estimate for SpeedLink is about $140 per RSH (Figure 3-2, above).
Justifications for the increments of operating cost per RSH, derived above and illustrated in Figure 3-2 (below), may be summarized as follows:
1.)    Baseline cost per RSH, derived from operation of standard buses ($100 per RSH).
2.)    Adjustment for higher cost per RSH owing to operation of articulated buses, which cost 25 percent more to operate per vehicle-mile than standard buses (additional $25 per RSH, holding the number of vehicle-miles per RSH constant).
3.)    Adjustment for higher number of vehicle-miles per RSH, holding fuel economy (miles per gallon) constant. Although fuel consumed / mile remains constant, miles / hour increases, and therefore fuel consumed / hour ( = fuel consumed / mile * miles / hr) also increases.
For the purpose at hand, the author used 15 revenue vehicle-miles per RSH for the "baseline," and 25 revenue vehicle-miles per RSH, implied by MAC operating-cost estimates, for SpeedLink. The implied fuel cost per RSH from 1.) above, rounded to the nearest dollar, was $6. The resulting "adjusted" fuel cost per RSH is $10, an increase (from 1.) of $4 per RSH. To avoid spurious precision, the resulting sum of $129 was rounded to two significant digits, giving $130 per RSH.
4.)    Adjustment for reduced fuel economy resulting from higher "cruise speed" between stops.
The annual revenue vehicle-miles per RSH for statistic for DDOT and SMART imply commercial (passenger) speeds in the range of 15-25 mph. These figures imply in turn "cruise speeds" of 25-35 mph. For the purpose at hand, the author used 25 mph as the "baseline" cruise speed. SpeedLink service standards and operating-cost estimates imply cruise speeds in the range of 35-45 mph. Results obtained by Sims and Miller imply that a cruise speed increase from 25 to 45 mph, with an average of two stops per mile will result in a 55 percent increase in fuel expense per mile. For the purpose of this adjustment, the number of vehicle-miles per RSH is constant. The author increased by 55 percent the "adjusted" fuel cost from 2.) above, $10 per RSH. The result, $15.50, represents an increase of $5.50 per RSH. The resulting sum, $135.50, was rounded to two significant digits, giving $140 per RSH.
4.) Additional Operating Expense: Underestimation of Service Required Relative to Ridership
Underestimation of service level with respect to ridership is a chronic problem associated with recent U.S. transit projects. The initial car fleets of several recent LRT systems proved to be too small for the traffic that developed after opening.
Table 4-1: Weekday Passenger Traffic – Peer Systems
"Unlinked Passenger Trips"
"Passengers Per Rev Hour"
"Passengers Per Rev Mile"
"Passengers Per Route Mile"
Light Rail
St Louis
San Jose
Los Angeles Metro Rapid Bus Lines
Line 720
Line 750
Notes: "Passengers Per Route Mile" presented by MAC refer to "directional" route-miles for LRT lines, and "bi-directional" route-miles for Los Angeles Metro Rapid Bus lines. This statistic was therefore doubled for the Rapid Bus lines.
"Passengers Per Rev Hour" and "Passengers Per Rev Mile" presented by MAC for LRT refer to revenue train-hours and train-miles, not vehicle-hours and vehicle-miles.
Source: Metropolitan Affairs Coalition, June 2001.
For the prospective SpeedLink corridors, ridership forecasts were developed based on 1.) an incremental logit model, and 2.) comparison with "peer" systems in terms of passengers per revenue-hour, revenue-mile, and route-mile. These three parameters were averaged; ridership forecasts in each category were based on 75 percent of the "peer" averages (Table 4-1, above).
Service was outlined to operate daily, from 4 am to 1 am, with:
--5-minute headways during weekday peak periods
--10-minute headways during weekday midday periods and on weekends.
--20-minute headways during evening hours.
For the Woodward corridor, the average of the four forecasts is 23,300 passengers per weekday. MAC does not provide annual ridership estimates, but does provide "Net New Passengers:" 1,655,000 for the Woodward corridor. The difference between "current" and "predicted" ridership is 4,900 (that is, 4,900 "new" passengers per weekday); 1,655,000 divided 4,900 equals 338 weekday equivalents. This number, times 23,300, implies 7,875,000 annual passengers.
Table 4-2: Passenger Traffic and Service Characteristics – Peer Systems
ARL, mi
Pass-Mi / Veh-Mi
Traffic Density
Weekday Service
Pass Speed, mph
Light Rail
St Louis
San Jose
Los Angeles Metro Rapid Bus Lines
Line 720
Line 750
Notes – LRT: "ARL," = Average Ride Length [per unlinked passenger trip], annual average derived from NTD statistics. Weekday ARL may be higher.
"Annual Pass-Mi / [revenue] Veh-Mi" derived from NTD statistics.
"Weekday Pass-Mi / [revenue] Veh-Mi" derived from MAC ("Unlinked Passenger Trip"), weekday service as shown in public timetables, and estimates of average train length.
"Traffic Density," passenger-miles per mile of route per weekday. Derived from MAC ("Unlinked Passenger Trip") and ARL. May be underestimated owing to use of annual average ARL.
"Weekday Service" the number weekday revenue vehicle-miles per mile of route. Derived from public timetables and estimates of average train length.
"Pass Speed (mph)" estimated from public timetables.
Sources: Final Report: Los Angeles Metro Rapid Demonstration Program, MAC, NTD.
MAC did not provide an estimate for average ride length per boarding (ARL) in the various projected SpeedLink corridors. This is an unfortunate oversight. ARL is an extremely important parameter, related directly to annual passenger-miles and therefore, traffic density and the service level required by a given ridership level.
DDOT carried (at 2000) an ARL of 4.5 miles, which is high for a large U.S. urban surface bus system. SMART carried an ARL of 6.7 miles at 2000. U.S. freeway-express, HOV, and rail lines typically carry a greater ARL than surface bus systems owing to higher passenger speed. Given the length of the Woodward corridor, the passenger speed implied by MAC (30 mph) and the location of major destinations along the corridor, the minimum plausible ARL for the Woodward corridor would be eight miles. The St. Louis LRT line, which was nearly half the length of the planned Woodward corridor at 2000, carried an ARL of 6.6 miles (Table 4-2, above). The implied annual ridership and the estimated ARL for the Woodward corridor suggest 63 million annual passenger-miles.
MAC states the annual service level for the Woodward corridor as 2,587,938 revenue vehicle-miles (111,496 RSH; 23.2 revenue vehicle-miles per RSH). This implies 24.3 annual passenger-miles per annual revenue vehicle-mile (Figure 3-2). Adjusted for differences in vehicle length, this figure exceeds that carried by all "peer" services except St. Louis LRT and Los Angeles Metro Rapid ("Rapid Bus") Line 720 (Wilshire-Whittier). A more realistic figure, based on actual "peer" system operation (Table 4-2, above), is 14-18 annual pass-mi per veh-mi. The "median" value (16 pass-mi per veh-mi), together with the passenger-mile estimate above, imply 3.9 million annual revenue vehicle-miles. This is 52 percent more service than was estimated in the MAC report.
The $140 per SpeedLink RSH estimated above implies an annual operating cost of nearly $24 million, more than double the $11 million estimated by MAC (Figure S-1, above).
5.)Additional Operating Expense: Omission of Station Operating and Maintenance Costs, Security Costs and Self Service Fare Collection Costs
MAC did not address fixed operating costs, including station operation and maintenance costs, security costs, and costs associated with self-service fare collection. This is a serious omission. Costs of this nature are, of course, reported by transit operators to the FTA and included in the annual average cost per RSH statistics found in the NTD.
Among the various SpeedLink attributes outlined by MAC are "Lighted and heated stations," "'Real-time' passenger information," "Ticket vending machines" and "Improved Customer Experience." Other attributes indicate clearly that SpeedLink would use self-service fare collection: "Multiple-Door Boarding/Alighting" and "Off-Vehicle Fare Payment."
Based on experience with surface LRT stations of straightforward design, overall cost for station lighting, cleaning and maintenance (including fare-collection equipment) is likely to approximate $100,000 per year. This implies $3 million per year for Woodward corridor station operating and maintenance. This estimate may be conservative given the specification of heated stations (or shelters). Staffing requirements for station cleaning and maintenance may be estimated based on two to three stations per full-time employee.
The public may demand a high security presence aboard SpeedLink vehicles, and at parking lots and stations. This, and the need for enforcement of self-service fare collection, suggests that SpeedLink security costs may be relatively high. Even with a single corridor, annual security and fare-enforcement costs may total up to $3 million during the initial years of operation, when the operator might reasonably choose to "invest" in a reputation for customer security. These costs are not likely to increase in proportion as the network expands.
The impact of an annual $6 million in fixed operating costs on SpeedLink unit operating cost depends, of course, on the amount of service (RSH) actually operated. Factors outlined above suggest an overall operating cost of $170-190 per SpeedLink RSH.
6.) Annual Operating Expense: SpeedLink and LRT
It is clear that other modes considered by MAC might cost less to operate in some corridors than SpeedLink, and that the cost advantage of other modes might increase as ridership grows. For the sake of brevity, the analysis below compares SpeedLink with one alternative, LRT, in one corridor, Woodward Avenue.
The average of the four Woodward corridor forecasts presented by MAC (23,300), multiplied by the minimum plausible ARL (eight miles) implies 186,400 passenger-miles per weekday. This figure, divided by the planned route length (25.3 miles), implies a weekday traffic density of 7,400 passenger-miles per mile of route per weekday.
MAC has forecast a moderate traffic density for its starter SpeedLink corridor (Figure 6-1, above). This is significant for three reasons: 1.) Increased traffic density, resulting from ridership growth, would require additional service, particularly during peak periods, 2.) traffic density changes substantially given incremental changes in weekday ridership or, in particular, ARL, and 3.) buses do not provide labor and infrastructure economies of scale.
Given a weekday ridership of 30,000 (which would result from five years of ridership growth at five percent per year) and an ARL of nine miles (similar to that carried by the LRT Blue Line in Los Angeles), the implied weekday traffic density is 11,000. This in turn implies that weekday service would have to be increased substantially, by up to 50 percent, in order to carry this level of ridership.
Labor and infrastructure economies of scale, provided by rail systems, do not arise from differences between average and marginal operating cost. Instead, they reflect system characteristics such as operation of multiple-car trains by a single person, and spreading of infrastructure costs over a larger quantity of service. This is demonstrated by the relationships among the seven "peer" LRT systems between increased service (annual vehicle-miles per route-mile) and reduced cost per annual RSH (Table 6-1, below).
Table 6-1: Service Level and Annual Operating Expense
Annual RSH per Route-Mile
Annual Operating Expense (2000; 2000$s)
per RSH
per Route-Mile
Light Rail
St Louis
1.4 million
1.6 million
2.3 million
1.1 million
San Jose
1.4 million
1.1 million
1.1 million
Los Angeles Metro Rapid Bus Lines
Line 720
No report
0.9 million
Line 750
No report
0.4 million
Notes– LRT: "Annual Operating Expense (2000) per RSH" for LRT refers to revenue vehicle-hours (not train-hours).
Notes – Los Angeles Metro Rapid Bus: "Annual Operating Expense (2000) per RSH" is LACMTA system average for 2000 (NTD).
"Annual Operating Expense (2000) per Route-Mile" based on annual RSH totals:
Line 720 – 224,000 RSH; Line 750 – 72,000 RSH.
Sources: Final Report: Los Angeles Metro Rapid Demonstration Program, NTD.
A Los Angeles budget document contains an "Activity Based Light Rail Cost Model" ("Proposed Budget For Fiscal Year Ending June 30, 2003;" no longer available online). This suggests that a one-percent increase in annual LRT RSH leads to a 0.5-percent decrease in operating cost per RSH.
Large increases in LRT service level may lead to substantial reduction in unit operating cost. During the mid-1990s, LRT lines in St. Louis and Portland had similar route length, similar service levels, and a two-car maximum train length. Both posted similar operating cost per RSH; adjusted for inflation, this figure fluctuated between $180-200 (2001$s). Then, Portland opened its "Westside MAX" LRT line and greatly increased service (per unit of route length). Inflation-adjusted operating cost per RSH, which averaged $200 during 1996-1998, plummeted to $140 (average during 1999-2001). A 30 percent reduction in operating expense per RSH given a substantial service increase is unlikely with buses – absent measures to reduce labor costs.
Economy of LRT operation is maximized when trains are lengthened as needed to accommodate traffic demand, instead of increasing the number of trains in service. Sacramento operates LRT service every 15 minutes throughout the day, but operates single cars, two-car trains or four-car trains as needed. Sacramento’s operating cost of $177 per RSH (200), significantly lower than the average for the seven "peer" systems (Table 6-1, above), reflected this. Among the "peer" systems, Baltimore’s operating cost of $167 per RSH reflected similar factors.
Considering differences in vehicle size, $170-190 per RSH for articulated buses is equivalent to $250-290 per RSH for LRT. This is considerably higher than the average for the seven "peer" LRT systems (Table 6-1, above).
Notes: Passengers per meter of vehicle length is a standard suggested by Parkinson and Fisher (1996) to place all systems and modes on an equal footing.
Median values exclude operators in Boston, Montréal, New York and Toronto.
Parity for bus and LRT in terms of service effectiveness – passenger-miles per revenue vehicle-mile – is difficult to justify. Data collected in the U.S. and Canada from the mid-1980s make clear that bus rapid transit (busway and HOV) services do not perform equally with rail services in terms of peak-period productivity. Rail systems (outside of Boston, Montréal, New York and Toronto) average about four passengers per meter of vehicle length. The figure for buses is lower, about three pass/m of vehicle length; the rail median is greater than the bus maximum. The fundamental underlying causal factor is, of course, consumer choice although clarification of details awaits additional research.
The fact that U.S. consumers tolerate higher levels of peak-period crowding aboard railcars than aboard buses is known to transit operators and is incorporated into service planning. For example, Sound Transit and King County Metro plan mixed operation of light rail vehicles and buses in the downtown Seattle transit tunnel (DSTT). Achievable-capacity estimates are based on 137 passengers per vehicle for LRT, and 46 pass/veh for articulated buses. The LRT figure (which implies 4.7 – 5.0 pass/m based on the planned vehicle length of 90-95 feet) reflects actual experience in nearby Portland. The bus figure (2.5 pass/m) is based on actual experience in Seattle. King County Metro has found that planning for more than 80 percent of seated capacity overall during peak periods usually results in periodic overloads, requiring some passengers to stand. Metro has also learned that regularly exceeding the 80 percent figure leads to a large number of passenger complaints. It has concluded that an average peak-period vehicle occupancy greater than the 80 percent figure may discourage ridership (Sound Transit and King County Metro 2001).
Another important factor is the strong relationship between weekday traffic density and service effectiveness for LRT (Appendix I, Table AI-2). This does not arise exclusively from increased peak-period crowding (as demonstrated by the estimated weekday service-effectiveness figures). Several recent LRT systems have experienced strong growth in weekend and special-event traffic. In some cases, weekend traffic has grown faster than weekday traffic. St. Louis is one city where LRT has demonstrated its capability to accommodate very high traffic volumes for special events: the current one-day record, 160,721 passengers (set on Saturday, July 1, 2000) was nearly four times greater than the weekday average (for 2000).
Table 6-2: Estimation of Woodward Corridor Operating Expense
Traffic Density
LRT Pass-Mi per Veh-Mi
LRT Cost per RSH
Annual Operating Expense
Articulated Bus
Light Rail
$29.2 million
$27.4 million
$1.8 million
37.3 million
30.1 million
7.2 million
45.3 million
31.5 million
13.8 million
53.4 million
32.4 million
21.0 million
61.4 million
32.8 million
28.7 million
Comparative annual operating costs, presented in Table 6-2 (above), were calculated based on:
--7,500 – 17,500 pass-mi per route-mi per weekday.
--338 weekday equivalents per year.
--25.3 mile line length.
--23.2 revenue veh-mi per revenue veh-hr.
--16 pass-mi per veh-mi for articulated bus.
--$140 per RSH for articulated bus, plus $5 million annual fixed operating costs.
--Pass-mi per veh-mi for LRT calculated by linear interpolation with reference to traffic density, based on St. Louis, Dallas, Cleveland, San Jose, Pittsburgh and Baltimore. Buffalo excluded for atypically low service effectiveness with reference to traffic density.
--Operating cost per RSH for LRT calculated by linear interpolation with reference to traffic density, based on St. Louis, Dallas, Cleveland, San Jose and Pittsburgh. Baltimore and Buffalo excluded for atypically low operating cost per RSH.
Use of 16 pass-mi per veh-mi and $140 per RSH for articulated bus at all traffic-density levels may lead to overestimation of articulated bus operating expense. However, these assumptions are conservative and defensible. Bus rapid transit facilities are rare in the U.S. and Canada, and published data for certain critical performance indicators are scarce. For example, no published data could be located for operating cost per RSH or annual service effectiveness of lines or systems using 100 percent articulated buses. The evaluation of the Los Angeles Metro Rapid demonstration, published by the Federal Transit Administration, presents marginal operating costs ("before" and "after"). As noted above, these are of little use for the purpose at hand. The very high service effectiveness of Metro Rapid Line 720 (Wilshire-Whittier) reflects characteristics unique to the corridor, particularly along Wilshire Boulevard. High levels of population, density, and employment along this 15-mile section give rise to high levels of transit ridership. Wilshire bus and rail services carry one-third of the weekday ridership carried by the entire DDOT system. This corridor also has a large "hidden" market for transit service (given sufficient travel-time savings and adequate capacity).
On the other hand, available information suggests strongly that parity between modes should not be assumed. Published data make clear that, in Ottawa, operating cost per RSH rose sharply and fuel economy declined as the busway ("Transitway") network was opened in stages and the operator deployed a large fleet of new articulated buses. Observed passenger behavior during peak periods suggests that consumers prefer to ride in the front section of articulated buses and avoid the rear section when possible. This is significant in light of the fare system (passengers may board at any door at Transitway stations) and the fact that the rear section of articulated buses provides a poorer ride quality than the front section (Vuchic 1981).
A traffic density of 7,500 passenger-miles per mile of route per weekday, together with the other factors outlined above, implies that SpeedLink and LRT operating costs would be almost identical (Figure 6-4, above) at this traffic level. This traffic density is slightly greater than that carried by Baltimore and Pittsburgh among the "peer" LRT systems, and implies about 24,000 passengers per weekday (given the minimum plausible ARL of eight miles).
However, given a weekday traffic density of 10,000, equivalent to that carried by Dallas LRT and implying about 32,000 passengers per weekday, the operating-cost difference becomes significant: $7 million per year. The operating-cost advantage of LRT widens to nearly $30 million per year given a traffic density of 17,500, equivalent to that carried by St. Louis LRT and implying about 55,000 passengers per weekday. If the ARL proves to be greater than eight miles, the weekday ridership associated with the various traffic-density levels would be less, and vice versa.
7.) Lower Operating Cost Offsets Greater Capital Investment
If various alternatives cost exactly the same to operate given an identical level of consumption (ridership), then selection of the alternative with the lowest capital cost follows in straightforward fashion. However, if operating costs are not identical, the difference may be sufficient to justify a larger capital outlay.
Table 7-1: Annual Operating Cost Difference and Incremental Capital Expenditure Offset
Annual Operating-Cost Reduction of:
Offsets Incremental Capital Expenditure of:
$0.5 million
$6.3 million
1.0 million
13 million
2.5 million
31 million
5.0 million
63 million
7.5 million
94 million
10 million
125 million
15 million
188 million
20 million
251 million
25 million
313 million
30 million
376 million
35 million
438 million
Incremental Capital Expenditure of:
Offset by Annual Operating-Cost Reduction of:
$10 million
$0.8 million
25 million
2.0 million
50 million
4.0 million
100 million
8.0 million
250 million
19.9 million
500 million
39.9 million
1,000 million
79.8 million
Note: Based on 7 percent discount rate and 30-year project life; calculated using "Engineering Economics Calculator," (2003.9.21).
It is true, for example, that an annual $8 million saving in operating cost would pay back $100 million in additional capital expenditure within 13 years. However, the annual savings would also pay back the opportunity cost of this capital, based on a 7 percent discount rate, within a 30-year project life. The additional investment would therefore be economically justified.
8.) Underestimation of Speedlink Capital Cost
The finding that MAC significantly underestimated service levels with regard to ridership (Section 5, above) suggests that MAC also underestimated SpeedLink capital costs. For the Woodward corridor, SpeedLink facilities and vehicles are estimated to cost $188 million, with an additional $30 million for the operating base (which would eventually serve additional SpeedLink corridors ). The total cost for the "starter" SpeedLink line works out to $8.6 million per mile.
The "peer" LRT systems carry between four and ten percent of two-way, all-day traffic during the busiest hour, in the busier direction (Table 8-1, below). The median is six percent, and only two of the seven "peer" LRT systems carry less than this. Therefore, it seems prudent to assume that the Woodward corridor would need to accommodate 6-10 percent of average weekday ridership, 1,400 – 2,300 passengers, during the busiest hour, in the busier direction.
Table 8-1: Peak-Period Passenger Volumes and Service Levels
Peak-Period One-Way Passenger Volume
Unlinked Passenger Trips
Peak / Total (percent)
Light Rail
St Louis
San Jose
Los Angeles Metro Rapid Bus Lines
Line 720
No Report
Line 750
No Report
Sources: MAC, "Peak-Period Vehicle Occupancy Statistics for U.S. and Canadian Rapid Bus and Rapid Rail Services."
As noted above, MAC has outlined weekday peak headways of five minutes, or 12 vehicles per hour (requiring 27 vehicles in service). This implies 117 – 192 passengers per vehicle (or 6.4 – 10.5 pass/m, assuming 60-foot vehicles). This is far above the median (2.9 pass/m) – and maximum (3.7 pass/m) – for busway and HOV services in the U.S. and Canada, observed post-1985. Based on realistic vehicle-occupancy levels of 2.5 – 3.0 pass/m (45-55 pass/veh) as observed in other U.S. cities, MAC would need to operate 25-50 vehicles per hour. Failure to provide this higher level of peak-period service would lead inevitably to ridership shortfalls.
In Los Angeles, the operator has learned that operation of two-minute headways with buses receiving traffic signal priority (as on Line 720) leads to bunching and overcrowding because consecutive buses cannot receive priority. The needs of cross traffic (which includes busy bus lines) must be considered. The minimum headway that would permit all buses to receive priority, based on the 90-second traffic signal cycle, is 180 seconds, or three minutes, implying a maximum of 20 vehicles per hour (White 2002). Although Line 720 is noticeably faster than previous limited-stop bus services, the observed peak-period passenger speed is 12-13 mph, far short of the implied 30 mph SpeedLink standard.
Operation of 25-50 vehicles per hour over the proposed SpeedLink preferential-lane facilities, while providing a 30 mph passenger speed, would be very difficult. Full (absolute) traffic signal preemption would impose intolerable delays on cross traffic. Additional investment would be required to permit large numbers of buses to pass through intersections without such disruption (e.g. underpasses or overpasses). This, in addition to the required larger fleet size, suggests that the starter SpeedLink corridor would cost much more to build than estimated. If the additional capital were not provided, then SpeedLink would not be able to accommodate the projected traffic, and would not be able to provide the projected service quality.
A prudent design standard for SpeedLink, allowing for future traffic growth, would reflect the maximum peak-period volume observed on the "peer" systems (2,500 pass/hr, St. Louis LRT), and actual U.S. experience with articulated buses (e.g. Seattle, 80 percent of seated capacity, described in Section 4 above). This implies that, in the Woodward corridor, SpeedLink should be designed for a maximum peak-period capacity equivalent to 55 articulated buses per hour. This, given the planned 30 mph passenger speed, would require the equivalent of a segregated busway. Such a facility would cost at least twice the amount estimated by MAC, or more.
As noted above, it is clear that MAC has underestimated the required vehicle fleet size for the Woodward corridor. The implied cost estimate for articulated, low-floor, hybrid-powered vehicles for SpeedLink is $800,000 – $900,000 each. It is pertinent to note that recent advanced-technology articulated vehicles for BRT services (e.g. Boston's Silver Line) are reported to be higher, up to $1 million. This is well within the cost range for LRT vehicles when differences in vehicle size and vehicle life are considered.
8.) Conclusions
The proposed SpeedLink facilities would cost substantially more to operate, and substantially more to build, than has been estimated. If the issues outlined above are ignored or papered over, it is likely that the SpeedLink proposal will lose vital credibility. This, in turn, would jeopardize prospects for successful implementation, not only for SpeedLink in Metro Detroit, but for similar projects elsewhere.
Additional research is indicated regarding operating costs and other issues related to bus rapid transit operation, such as service levels requisite with various peak-period traffic levels. Obviously, this research needs to be conducted prior to undertaking a BRT project on the scale planned by MAC. "Think rail, use buses" makes a catchy slogan, but may prove technically and financially impractical for major, high-density urban travel corridors. To its credit, MAC obviously recognizes that all service attributes must equal those provided by rail transit in order for SpeedLink to succeed.
Even if bus rapid transit is found to cost no less to build and significantly more to operate than rail in major urban corridors – in other words, if the worst-case scenario were proven true – BRT would retain the very significant advantage of incremental implementation. Preferential measures, reserved lanes and segregated alignments can be started as short, unconnected segments, to be expanded and connected as traffic justifies and finances permit. This is not an option for rail development. Another important advantage is that buses can be purchased, borrowed or transferred much more quickly than additional railcars can be procured.
Even if rail transit was proven to be more cost-effective in the corridors of highest demand, it is not likely that all eleven corridors outlined by MAC for Metro Detroit would generate traffic sufficient to justify the investment in rail facilities. Thus, construction of SpeedLink facilities in some corridors would be justified. For this reason alone, it should be clear that the credibility of the SpeedLink proposal should not be compromised. It is hoped that the issues outlined above will not be ignored.
9.) Epilogue
At 2007, the MAC SpeedLink plan had effectively been shelved. Attempts to create a regional transit authority had not succeeded, and no agreement on major transit investments had been reached.
In 2002, the Michigan legislature approved a bill to create a Detroit Area Regional Transportation Authority (DARTA), but this was vetoed by Governor John Engler. A similar bill was introduced during the 2003 session but was not approved. Later that year, Metro Detroit officials and Governor Jennifer M. Granholm crafted a plan to create DARTA with an "Intergovernmental Agreement." This agreement, among the City of Detroit, the Suburban Mobility Authority for Regional Transportation and the the Regional Transit Coordinating Council (RTCC), was challenged in court by the American Federation of State, County and Municipal Employees (AFSCME). The Wayne County Circuit Court ruled in 2004 that, while RTCC did not have the power to enter into the agreement, it remained in force without RTCC and DARTA therefore remained in existence. The Michigan Court of Appeals upheld this ruling in 2005. DARTA has received federal funds for "start-up" activities such as hiring of staff and beginning of planning activities. However, a 2005 article in Metro Times Detroit described DARTA as "about two steps away from existing exclusively on paper" ("Fixing a hole," 6/1/2005, ).
Funding is a major problem that hinders public transit development in Detroit. A dedicated, voter-approved regional tax, of the sort used in other cities, is not permitted under the Michigan Constitution. The SpeedLink construction cost estimate, $2 billion over 25 years, was criticized as too expensive.
RTCC, which was created by the legislature in 1964 to plan and implement public transportation policy in Southeastern Michigan, began a low-key "Detroit Regional Mass Transit Initiative" in 2007. A tentative "vision document" outlined an initial "rapid transit" corridor on Woodward, between downtown and New Center, three miles. The woodward corridor would be extended and four additional corridors added during subsequent phases. RTCC planned to retain consultants to recommend a detailed transportation plan, financing – and choice of mode, light rail transit or bus rapid transit.
The Detroit Department of Transportation (DDOT) hosted four public meetings in July 2007 for three potential "rapid transit" routes, on Woodward Avenue, Michigan Avenue and Gratiot Avenue. The meetings were part of the Detroit Transit Options for Growth Study (DTOGS) that was planned as the initial stage of application for a Federal Transit Administration grant. Modes planned for evaluation were "Bus Rapid Transit (BRT)," "Light Rail Transit (LRT)/Modern Streetcar," and upgraded conventional bus service.
The authors express sincere appreciation to Lyndon Henry, E. L. Tennyson, PE, Van Wilkins and two other individuals who provided useful input and feedback during preparation of this paper.
Articulated Bus Report. 1982. Cambridge Systematics, Inc. (Cambridge, MA). Prepared for Office of Service and Management Demonstration, Office of Technical Assistance, Urban Mass Transportation Administration, U.S. Department of Transportation; DOT-TSC-UMYA-82-22.
Articulated Buses - A Planning Handbook. 1984. Office of Service and Management Demonstration, Technical Assistance Program, Urban Mass Transportation Administration, U.S. Department of Transportation, DOT-TSC-1752.
Connecting the Past - Creating the Future. [Detroit Transit Options for Growth (DTOG) Study].
DARTA - the Detroit Area Regional Transit Authority (2007.9.22).
"Engineering Economics Calculator" (2007.9.22).
Evaluation of Joint Operations in the Downtown Seattle Transit Tunnel. 2001. Seattle: Sound Transit and King County Metro.
Final Report: Los Angeles Metro Rapid Demonstration Program. 2001. Prepared by Transportation Management & Design, Inc., for Federal Transit Administration, July 2001.
ht tp://www. (This report is no longer available online).
Mass Transit: Bus Rapid Transit Shows Promise. 2001. Washington, DC: U.S. General Accounting Office, GAO-01-984, September 17, 2001.
Metro Times Detroit. (2007.9.22).
Motown Tranzit. (2007.9.22)
National Transit Database, Federal Transit Administration, U.S. Department of Transportation,
Parkinson, Tom, and Ian Fisher. 1996. Rail Transit Capacity (Transit Cooperative Research Program Report No 13). Washington, DC: Transportation Research Board, National Research Council.
Proposed Budget For Fiscal Year Ending June 30, 2003. 2002. Los Angeles: Los Angeles Metropolitan Transportation Authority (this document is no longer available online).
Sims, D., and Eric J. Miller. 1982. "Energy Consumption of Alternative Fixed-Route Transit Modes," RTAC Forum, volume 5, number 1, 1982. (Tables from this study, in metric units, are presented in Meyer, Michael D., and Eric J. Miller. 1984. Urban Transportation Planning: A Decision-Oriented Approach. New York: McGraw-Hill.
SpeedLink: A Rapid Transit Option for Greater Detroit. 2001. prepared by Transportation Management & Design, Inc., for Metropolitan Affairs Coalition, Detroit, MI, June 2001.
Transportation Riders United. (2007.9.22).
Vuchic, Vukan R. 1981. Urban Public Transportation - Systems and Technology. Englewood Cliffs, NJ: Prentice-Hall.
White, Ron. 2002. "L.A. In High Gear With Bus System." Busline, September/October 2002.
Document History
Revised: November 6, 2002.
February 12, 2003.
Minor corrections and restoration of Figure S-1: September 27, 2003.
Epilogue added, links checked: September 22, 2007.
Appendix 1: Fleet Size and Transit Bus Fuel Consumption
The FTA website has a table comparing bus energy consumption with fleet size (ht tp://ww; inactive link at 2007.9.22, scroll down to Table 3-20). The following is condensed from the FTA table:
Table A1-1: Transit Bus Fleet Size and Fuel Consumption
Fleet Size
Fuel Consumption
1,000 or more
3.2 mpg
500 – 999
250 – 499
100 – 249
50 – 99
Below 25
All Motor Buses
Source: UMTA Section 15 data for 1989
Differences in fuel consumption rates arise from differences in vehicle size, and service characteristics such as cruise speed and average number of acceleration cycles per mile. It is not likely that fleet size per se is the determining factor.
Appendix 2: Transit Bus Energy Efficiency vs. Operating Speed
The energy efficiency achieved by transit buses increases significantly with cruise speed – up to a point, and particularly when stops are infrequent. Beyond this point, bus fuel consumption increases rapidly with operating speed – and the greater the stop frequency, the faster the rate of increase.
Performance statistics for individual bus lines in Los Angeles ("Quarterly Performance Report," Los Angeles: Los Angeles County Metropolitan Transportation Authority; note that these are no longer produced) suggest that bus operating cost per vehicle-mile decreases as "passenger speed" increases – up to about 15 mph. There is no clear trend for cost per RSH to 15 mph, but above this passenger speed, cost per RSH rises rapidly.
Analysis of NTD (1999) statistics (for 1999) reveals the following:
1.) Energy consumption per vehicle-mile falls significantly as "actual vehicle-miles" per "vehicle hour" (used instead of “revenue vehicle-miles / RSH to reduce uncertainty) increased from about 8 to about 15-17.
2.) A "gray area" exists where there is no clear trend. This extends roughly between 12 and 18 actual vehicle-miles / vehicle-hours.
3.) Energy consumption per veh-mi increases above 20 veh-mi / veh-hr, although the rate of increase varies substantially with stop frequency.
4.) All-bus operators which operate a significant percentage of service on freeways, busways or HOV facilities report lower energy consumption per vehicle-mile than those with a large percentage of surface, mixed traffic operation. The difference is on the order of 30 percent.
Item 1.) above implies:
2.7 mpg at 10 veh-mi / veh-hr.
4.5 mpg at 15 veh-mi / veh-hr.
The latter figure suggests the “typical” planning assumption of 4 mpg for full-sized transit buses.
Item 3.) above implies, for services operating above 20 veh-mi / veh-hr using standard buses and making infrequent stops:
4.8 mpg at 20 veh-mi / veh-hr.
4.2 mpg at 30 veh-mi / veh-hr.
These figures also suggest the "typical" planning assumption of 4 mpg, but are based on freeway-express services. Operationally, these bear little resemblance to the "typical" LRT service, with stops every 1-2 miles and a 45-55 mph cruise speed.
Item 3.) also implies, for services operating above 20 veh-mi / veh-hr, using full-sized transit buses and making relatively frequent stops:
3.4 mpg at 20 veh-mi / veh-hr.
1.9 mpg at 30 veh-mi / veh-hr.
Item 4.) above should be no surprise to motorists. Autos get better gas mileage on highways (higher average speed; fewer stops per mile) than in surface traffic. Many auto owners say in addition that cars used primarily for "highway" driving incur lower maintenance costs than those used exclusively for "city" driving.
Several caveats should be kept in mind. First, the "veh-mi / veh-hr" figures in this appendix are not directly comparable to the "cruise speed" figures presented by Sims and Miller. For the same service or system, cruise speed will be greater than the ratio of annual revenue vehicle-miles to RSH. Second, it is difficult to determine average distance between stops for bus systems using NTD data. Site-specific factors such as topography and climate may have a significant impact on transit bus fuel economy.
Appendix 3: Explaining the Findings of Sims and Miller
The findings of Sims and Miller reflect fundamental principles of physics, in particular, those related to the differences between rubber-tired vehicles powered by internal-combustion engines, and rail vehicles powered by electricity. These include:
1.) Differences in rolling resistance between rubber tires and steel wheels.
2.) Differences in energy efficiency, with respect to operating speed, between internal-combustion engines and electric motors.
--The relationship between cruise speed and energy required to accelerate to this speed is not linear. It is equal to 1/2 times the weight (mass) of the vehicle times the square of the speed (velocity). In other words, acceleration from 0 to 30 mph requires four times as much energy as acceleration from 0 to 15 mph.
Once the cruise speed has been reached (assuming a straight and flat road or track), the energy consumed by the engine or motors goes to overcome air resistance, internal friction, wheel-to-surface friction, and rolling resistance.
--Rolling resistance is not "friction." It arises because wheels – whether steel or rubber-tired – distort from true round under the weight of the vehicle (and its load). Rolling resistance may be visualized as an invisible "hill" that the vehicle must climb, even when operating along a level road or track. The greater the rolling resistance, the steeper the "hill." The coefficient of rolling resistance for rubber-tired vehicles is ten times greater than for rail vehicles (Vuchic 1981).
For an articulated transit bus carrying a full seated load, based on the equations derived by Vuchic, the energy required to accelerate to 15 mph is 1/37th of the energy required to overcome friction, rolling and air resistance for one mile at that speed. However, the energy required to accelerate to 30 mph is 1/11th of the energy required to overcome resistance for one mile. The energy required to accelerate to 45 mph is 1/5th of the energy required to overcome resistance at that speed. It is obvious that frequent acceleration cycles to a high cruise speed will have a strong upward influence on bus operating costs per RSH.
For an articulated light-rail vehicle carrying a full seated load, the energy required to accelerate to 15 mph is more than twice that consumed by the bus – which has 75 percent of the seating capacity. This remains true at 45 mph.
The bus consumes 18 times as much energy per mile in order to overcome resistance at 15 mph, but this difference narrows as cruise speed increases.
However, given a 45-mph cruise speed, the energy required to accelerate the railcar is more than three times that consumed by resistance for one mile. In other words, if the vehicle were accelerated to 45 mph and the power was then shut off, after one mile it would remain in motion at considerable speed. This permits rail vehicles to coast between stops, which may result in significant energy savings.
Given a 25-mile route length, an average stop spacing of 0.8 mile, a cruise speed of 45 mph, and adjusting for differences in vehicle size (i.e. one railcar = 1.25 articulated buses), articulated buses would consume up to three times as much energy per seat-mile as LRT. If this fact is ignored during analysis of alternatives, bus operating costs will be greatly underestimated with respect to rail costs.
--An electric motor can accept a short-term "overload" without damage, or much loss of efficiency. Therefore, an electric vehicle can accelerate faster than an internal-combustion vehicle having the same horsepower rating.
--Over the range of operating speeds and stop frequencies typical of U.S. urban and suburban rail transit, power consumption per hour by electric railcars is essentially constant. That is, identical vehicles will consume the same amount of power per hour, whether operated in relatively low-speed service with frequent stops, or higher-speed service with less frequent stops.
--High-performance electric rail vehicles provide faster acceleration, permit higher "cruise speed" between stops (and therefore faster running time) – and also consume more power. This arises not from faster acceleration, but from the higher cruise speed permitted by faster acceleration. Potential savings from faster operation, including a reduction in the number of cars, trains and staff required to operate a particular service, may offset the additional energy expense. This has been documented in cases where old, slow-accelerating rapid-transit cars were replaced with new high-performance stock. Philadelphia’s Market-Frankford Subway-Elevated line is a good example.
--Unlike an electric vehicle, the energy consumed by an internal-combustion vehicle may vary considerably with changes in acceleration rate. This arises because the operating efficiency of an internal-combustion engine varies considerably with load and operating speed (revolutions per minute). Faster acceleration may require operation at higher load and rpm than would provide the greatest fuel economy. Higher acceleration rates may also lead to greater torque-converter losses. Motorists know that a moderate rate of acceleration burns less gasoline than a "jackrabbit start" to the same speed.
--The large energy-efficiency difference between bus and LRT reported by Sims and Miller, for services providing high cruise speed between frequent stops, reflects differences in "thermal efficiency." The most efficient diesel engines have a "thermal efficiency" of roughly 40 percent. The comparable figure for LRT motors, including losses inherent in the control system, is roughly 90 percent. For each acceleration cycle, the diesel vehicle will consume 2 to 2.5 times more energy than an electric railcar based on this factor alone. Therefore, as the number of acceleration cycles increases, total bus energy consumption increases faster than for LRT under identical operating conditions.
It is true that the difference between modes narrows considerably if the thermal efficiency of the power plant is taken into account. However, power plants operate with greater efficiency than internal combustion engines in motor vehicles, and electric railway technology minimizes energy losses owing to rolling resistance, acceleration and other factors. The energy efficiency of electric power generation relative to motor vehicles has little influence over transit operating expenses, except to the extent that local prices for diesel fuel and electricity reflect this.