“So what’s the problem with BART?” (Martin H. Duke, April 6, 2016).
With reference to Metrorail, BART does better than Duke implies. It achieves significantly lower unit operating costs than Metrorail, and does much better in terms of energy efficiency and labor productivity. Granted, BART has “issues,” but does do some things right.
The exercise below is based on statistics for the 2014 fiscal year. This was the most recent year for which National Transit Database (NTD) “transit profiles” were available at the time of writing.
Three factors – at very least – complicate comparisons between BART and Metrorail. It’s true that the systems were about the same size (103.7 vs. 106.4 route-miles, during FY 2014). However, BART provides a higher average commercial speed. The “average” BART passenger traveled twice as far as the “average” Metrorail passenger (14 vs. 7 miles). A third issue – which sometimes gets lost in the shuffle – is differences in transfer rates. Fewer than four percent of BART’s “average weekday passengers” transferred between trains, but more than 20 percent of Metrorail passengers did. These differences have the effect of skewing certain performance measures – some but not all in favor of BART. “Average travel distance” (ATD) also raises significant system-capacity issues.
Detailed study of transit performance measures is neither simple nor straightforward. Definition of terms, uncertainty (i.e. tolerance) associated with statistics and appropriateness of comparisons are among the problems that might ensnare the unwary.
One should take care to avoid “spurious precision.” The stated tolerance of Federal Transit Administration (FTA) data is ±10 percent, at the 90 percent confidence level (i.e. given a random selection of data items, nine out of ten will fall within 10 percent of the “actual” value). This implies, among other things, that a percentage difference smaller than 10 percent might not be significant.
The following example uses what UITP (Union internationale des transports publics; International Association of Public Transport) describes as the “fundamental” transit performance measure: annual operating and maintenance (O&M) cost per annual passenger-kilometer (or mile).
BART posted $0.32 in O&M cost per annual pass-mi for FY 2014, while Metrorail posted $0.63. BART’s “score” on this and other indicators below = (BART statistic) / (Metrorail statistic); 51 percent in this example. In other words, BART’s cost per annual passenger-mile of travel (PMT) was half that of Metrorail’s.
Also of interest: BART’s annual O&M cost per PMT has remained virtually stable for 35 years – while ridership increased by a factor of 2.6. The inflation-adjusted average was $0.37 per PMT, and the statistic for most years during 1981-2014 was in the range $0.33 – $0.41. The implied tolerance was ± $0.04, or ±11 percent. This is not statistically significant (because of uncertainty introduced by adjusting for inflation).
(Various facts herein might not sit well with rail critics in general and “BART bashers” in particular, but that is not germane to the matters at hand.)
BART vs. Metrorail (FY 2014): Operating Statistics
–Infrastructure and equipment
Rt-Mi: Route-miles (system length): 97 percent.
CBD Tracks: Nominal capacity through the central business district (CBD): 2 vs. 6 tracks; 33 percent.
VAMS: Vehicles available for maximum (i.e. peak-period) service, annual average: 65 percent.
–Labor and energy inputs
O&M Work hours: Annual total (i.e. full-time and part-time) operating employee work hours: 42 percent.
kWh: Annual energy consumption, kilowatt-hours: 57 percent.
Fare revenue: Annual passenger revenue: 70 percent.
O&M cost: Annual total operating and maintenance cost: 56 percent.
O&M labor cost: Operators Wages + Other Salaries and Wages + Fringe Benefits: 54 percent.
–Service production (output; supply)
Veh-hr: Annual revenue vehicle-hours: 60 percent.
Veh-mi: Annual revenue vehicle-miles: 87 percent.
VOMS: Vehicles operated in in maximum service, annual average: 61 percent.
Boardings: Annual boardings: 47 percent.
Person-trips: Annual passenger journeys (passenger-trips, linked trips): 57 percent.
AWR-B: Average weekday boardings: 46 percent.
AWR-P: Average weekday person-trips: 55 percent.
Pass-mi: Annual passenger-miles: 109 percent.
Rt-mi: The BART and Metrorail systems were about the same size during FY 2014: 103.7 vs. 106.4 miles.
CBD Tracks: The system length statistic belies the fact that system configuration is highly dissimilar: Metrorail has three trunk routes through central DC, but BART has only one through downtown S.F. In other words, Metrorail has a 6 to 2 advantage in terms of nominal CBD track capacity. Also: “The other Washington” lacks significant natural barriers – which the Bay Area has in abundance (e.g. San Francisco, San Pablo and other bays, together with a mountain range or two).
VAMS: BART reported an average of 662 VAMS. Metrorail reported 1,016.
O&M work hours: BART reported 5.2 million O&M work hours. Metrorail reported 12.6 million.
kWh: BART purchased 290.6 million kWh. Metrorail purchased 510.7 million. Note that this statistic incorporates train lighting, heating and air conditioning, together with other uses of “traction” current.
Fare revenue: Annual passenger revenue was $416.6 million for BART, and $593.3 million for Metrorail.
O&M cost: $533.6 million for BART, and $952.6 million for Metrorail.
O&M labor cost: was $396.8 million for BART, and $737.3 million for Metrorail.
Re. O&M cost and O&M labor cost, above: I did a double-take, several times over, but these are the figures reported to FTA and entered into the NTD.
Veh-hr: BART operated 1.8 million revenue vehicle-hours. Metrorail operated 3.0 million.
Veh-mi: BART operated 64.8 million revenue vehicle-miles. Metrorail operated 74.1 million.
Re. Veh-hr and Veh-mi, above: Metrorail has a long-standing peak-period service constraint which the system is now working to fix: platforms are long enough for eight-car trains, but power supplies, yards and shops are not adequate. About 25 percent of peak-period trains have 8 cars, and management plans to increase this to 100 percent by FY2021. By contrast, the capacity issues facing BART will be much more difficult – and costly – to mitigate. Thus, one might expect BART’s “score” on this indicator to decline, relative to Metrorail, during successive years.
The ratio of Veh-mi to Veh-hr differs significantly between operators (see below), and this has the effect of skewing certain performance measures.
VOMS: BART operated an average of 534 vehicles during intervals of “maximum service. Metrorail operated an average of 878.
Boardings: BART reported 125.8 million annual boardings, while Metrorail reported 269.5 million. Both are down from FY 2013: BART by nearly one percent and Metrorail by nearly two percent. Both changes are “significant” in light of multi-year trends.
Person-trips: The difference in (implied) weekday transfer rates suggests the comparison between annual passenger journeys. Estimated values are 120 million for BART and 210 million for Metrorail.
AWR-B: BART carried 417,000 average weekday boardings, while Metrorail carried about 915,000.
AWR-P: BART carried 399,000 average weekday passenger journeys, while Metrorail carried 722,000.
Re. Boardings, Person-trips, AWR-B and AWR-P, above: the disparity in scores arises from the difference between systems in transfer rates. Fewer than four percent of BART AWR-Ps transferred between trains, but more than 20 percent of Metrorail AWR-Ps did.
Re. AWR-B and AWR-P, above: “Average weekday ridership” (AWR) is an important indicator – and one of the easiest to scramble. That’s because 1.) FTA requires various operators to report “boardings” for NTD purposes, 2.) FTA no longer breaks out AWR by mode, and 3.) statistics published by individual operators do not always conform to FTA standards. A very good example of 3.) is the historic AWR data distributed by the Washington Metropolitan Area Transit Authority (WMATA). These numbers are – explicitly – station entries (aka “linked trips,” “revenue passengers” or “originating passengers”). This, by the way, is how “most” passenger rail operators worldwide report traffic statistics. Whether or not one believes that transfers between trains should be counted, few would argue against the need for uniformity (which is a major reason why FTA data are so useful, and popular, for comparisons among U.S. operators).
The NTD statistic for BART AWR (417,286) is not compatible with the Metrorail statistic presented by WMATA (721,804_. So, I took the NTD statistic for WMATA AWR by “all modes” (nearly 1.4 million), then subtracted the “Metrobus” statistic (source: WMATA website). Result: more than 927,000. I then subtracted 7,000 to account for “MetroAccess” paratransit ridership. Result: 920,000 average weekday boardings for Metrorail.
I checked the Metrorail AWR estimate using data from the American Public Transportation Association (APTA) website. I averaged the reported AWR figures for Metrorail for 2013 third quarter (2013 Q3), 2013 Q4, 2014 Q1 and 2014 Q2 (that is, for for the four calendar-year quarters included in the 2014 fiscal year). Result: nearly 906,000. Computing an “arithmetic mean” is one way of reconciling the difference, but I decided to use “significant digits.” Result: AWR 915,000 for Metrorail.
One final check: the ratio of annual boardings / AWR is described as “number of weekday equivalents.” The planning “benchmark” for U.S. heavy rail is 300. A significantly larger value implies that the system attracts relatively less weekend traffic than “typical.” A significantly smaller value implies relatively more weekend traffic than “typical.” (Historic examples greater than 365 are known, and date from many decades ago when Saturday was the busiest day of the week for transit.) Implied values are 295 for Metrorail and 301 for BART. The difference (about two percent) is too small to be significant.
(I believe in transparency, thus the above explanation of “where this number came from.”)
Pass-mi: BART reported nearly 1,700 million pass-mi, while Metrorail reported more than 1,500 million. The BART statistic increased slightly from FY 2013, while the Metrorail statistic decreased by more than two percent. These changes are not significant without reference to multi-year trends.
To be continued . . .