Does A Mile In A Car Equal A Mile On A Train? Exploring Public
Transit's Effectiveness In Reducing Driving
Dr. John Holtzclaw
If you quit your car for the subway, your mile in the sub equals the mile not driven,
right? Wrong? Maybe? How well transit improvements reduce driving and vehicle miles
traveled (VMT) is poorly understood. The supposition that one passenger-mile (p-m)
substitutes for one VMT cannot be made. Major transit improvements could well alter
destinations, walking/bicycling/transit/driving choices and routes, and land uses.
If transit riders drive to the station, one argument goes, the total trip length might
be longer than the auto trip it replaces.
If transit facilitates and encourages denser mixed-use development, however, shortening
average trip lengths and allowing more walking and bicycling, as well as transit, that
passenger-mile on transit might "replace" more than one mile of auto use. One
would make as many, but shorter, trips due to the synergism of public transit and high
density mixed-use communities. The author's residence in the dense Chinatown-North
Beach-Nob Hill-Russian Hill neighborhood of San Francisco, for instance, has ample transit
service and over 700 restaurants, and probably as many local markets, within an
interesting and pleasant one-mile walk.
Some studies of this have compared urban areas with and without rail transit, some
compared neighborhoods within urban areas or within a variety of urban areas, and one
simulated alternative future based on travel behavior.
The results of these studies show a potential for strong transit leverage effects,
wherein one additional passenger-mile on transit results in reductions in driving
substantially greater than one mile. There is scatter in the transit leverage
measurements, but all are in the same direction. Additionally, larger transit leverages
accompany greater densification in land uses and longer operation of transit service. In
every case the transit leverage is higher than 1, usually much higher. The data are not
adequate to prove cause -- more research is needed.
2. Previous Studies
Pushkarev and Zupan in their pioneering 1980 study compared the six American regions
with rail transit (New York-northeastern New Jersey, Chicago, Philadelphia, San Francisco,
Boston and Cleveland) to other U.S. urban areas over 2 million population, concluding that
for every p-m ridden on transit, four vehicle miles were not driven.1
This supports the thesis that good rail transit encourages land uses which shorten
average trip lengths. This study used the U.S. D.O.T.'s 1974 National Transportation
Report Urban Data Supplement which covered whole urban regions, thus mixing data for
sprawling suburbs with the older, denser, more convenient central city areas within each
region. Since older central cities, even those without rail transit, developed at higher
densities than recent suburbs, this study might undercount the potential transit leverage
available from building dense, transit-served communities rather than sprawling suburbs.
John Holtzclaw's 1991 study of five Bay Area communities reached similar conclusions.
It found that every transit p-m ridden by San Franciscans reduced VMT by eight compared to
VMT and transit use in suburban Danville-San Ramon.2
Much of San Francisco developed prior to the enactment of zoning to limit density and
commercial operations in residential areas and mandate generous parking. It not only had
an 8 times higher net residential density and a 12 times higher gross population density
than Danville-San Ramon, it also had 21 times higher commercial density. Holtzclaw argued
that most San Franciscans' trips to visit friends and family, to restaurants, markets or
jobs were much shorter. Consequently, not only is the average automobile or transit trip
shorter, but walking or biking substitute for many vehicular trips.
Holtzclaw also found that in the 15 years since BART began service, Walnut Creek had
densified, and one mile its residents rode on transit (mostly BART) to reduce VMT by four,
compared to Danville-San Ramon. Prior to BART, both areas had developed as freeway
3. Transit Leverage For Newman-Kenworthy's World Cities
Peter Newman and Jeffrey Kenworthy toured the world compiling demographic, land use,
auto and public transit data for 32 major cities. 3
They surveyed five American cities with no rail public transit in 1980: Denver,
Detroit, Houston, Los Angeles and Phoenix; five with rail transit systems: Boston,
Chicago, New York, San Francisco and Washington, DC. They also surveyed Toronto, Canada;
five cities in Australia: Adelaide, Brisbane, Melborne, Perth and Sydney; and twelve
European cities: Amsterdam, Brussels, Copenhagen, Frankfurt, Hamburg, London, Munich,
Paris, Stockholm, Vienna, West Berlin and Zurich. All the cities outside the U.S. had rail
transit systems. Moscow, Tokyo, Hong Kong and Singapore, less like American cities, were
excluded from this analysis.
The data show that metropolitan area residents of non-rail American cities averaged
9,386 Vehicle Kilometers Traveled (VKT) per capita in 1980 and rode 251
passenger-kilometers per capita. Metropolitan area residents of American rail transit
cities drove 7,050 VKT/capita and rode 1,050 p-k/capita. Metropolitan area residents of
the Canadian, Australian and European cities drove 3,871 VKT/capita on the average and
rode 1,644 p-k/ capita on the average. Combining U.S. rail cities with the non-U.S. cities
gives 5,316 VKT/capita and 1,374 p-k/capita. All averages are population weighted.
Comparing driving and public transit use of non-rail American cities to American cities
with rail transit yields a transit leverage of 2.9. Comparing driving and public transit
use of non-rail American cities to American, Canadian, Australian and European cities with
rail transit yields a transit leverage of 3.6.
4. Transit Leverage For San Francisco Bay Area Communities
The author recently analyzed 28 communities in the San Francisco, Los Angeles, San
Diego and Sacramento areas, including the three previously analyzed in the San Francisco
As in the 1991 study, these three communities were selected because they contrast older
central city with newer suburban areas, and because good transit data is available on
them. In the 1994 analysis, the study area boundaries for Walnut Creek are smaller and San
Ramon is included alone. The density, transit service, local shopping, pedestrian
accessibility and VMT/capita for these communities are shown in Table 1. The density is
measured as households per residential acre, using 1990 U.S. Census and the Association of
Bay Area Governments (ABAG) data.
The measure of transit service is the daily average number of transit vehicles per hour
on weekdays within a 1/2 mile walk from rail or ferry stations, or a 1/4 mile walk from
bus stops. These are averaged for the whole community. Transit vehicles are standardized
to 50 seats. Data is from local planning departments and transit agencies.
Neighborhood shopping is the fraction of the community's households with four specific
local commercial establishments (a mix of food markets, drugstores or restaurants) within
a 1/4 mile walk. Data is from local planning departments or surveys by the author.
The pedestrian accessibility is a combination of factors: fraction of local streets
that extend through to major streets (as opposed to dead-end streets with no walkways
through to the next street), fraction of roadway below 5% grade, fraction of streets with
sidewalks, convenience of the building entry to the sidewalk (average building setback),
and fraction of streets with traffic controlled by stop signs or lights at least every 600
feet. The data is from street, zoning and topological maps, zoning codes, and local
planning and public works departments.
The VMT/capita is the product of the auto ownership per capita and the average
VMT/vehicle for the community. The auto ownership is from the 1990 U.S. Census.
VMT/vehicle was calculated by the California Bureau of Automobile Repair using odometer
readings for cars undergoing mandatory biennial smog-checks.
The passenger-miles on transit were calculated using the same methodology as the 1991
4.1 San Francisco Transit Passenger-Miles
The annual transit p-m for San Franciscans is assumed to be the sum of San Francisco
Municipal Railway p-m plus San Franciscans' p-m on BART. Muni estimates 459 million total
annual passenger-miles, or 634 p-m per capita at a 1990 population of 724,000.5
This total includes some residents of adjoining counties, as well as visitors to the
City. However, some San Francisco residents ride the suburban buses. Since Muni totals
many more passenger miles than these other transit systems, and San Francisco is the
central transit destination, it is likely that more passenger miles are put on Muni by
non-San Franciscans than San Franciscans put on other bus systems. Ascribing all 459
million passenger miles on Muni to San Franciscans and ignoring San Franciscans' miles on
these other transit systems probably overestimates San Franciscans' passenger miles,
decreasing the size of the calculated transit leverage.
Calculations from the BART 1992 ridership data and their 1992 passenger survey indicate
that San Franciscans rode 486,575 p-m on the average weekday.6
Since BART carries 165,000 passengers on a average weekend, compared to 255,000 on
weekdays, the average annual BART ridership by San Franciscans is 198 p-m/capita.7
Adding Muni and BART passenger miles gives 832 annual transit p-m, see Table 1, which
is likely to be a slight overestimate of San Franciscans' transit use.
4.2 Walnut Creek Transit Passenger-Miles
Walnut Creek transit passenger miles were calculated from BART data for BART ridership
and for buses serving it. BART and Central Contra Costa Transit Authority (CCCTA) operate
buses connecting to BART stations. BART's data indicates that the 60,400 Walnut Creek
residents traveled 88,395 p-m on BART daily, giving 861 p-m/capita annually. Similarly,
Walnut Creek residents traveled 1770 p-m daily, or 9 p-m/capita annually on BART and CCCTA
buses serving BART stations. The Walnut Creek total is 870 p-m/capita annually for all
4.3 San Ramon Transit Passenger-Miles
San Ramon transit passenger miles were similarly calculated from BART data for BART
ridership and for buses. BART and CCCTA operate buses connecting to BART stations. BART's
data indicate that the 35,100 San Ramon residents traveled 18,480 p-m on BART daily,
giving 153 p-m/capita annually. Similarly, San Ramon residents traveled 3242 p-m daily on
buses serving BART, or 29 p-m/capita annually on BART and CCCTA buses serving BART
stations. Similarly, San Ramon residents rode CCCTA buses not going to BART 1 p-m/capita
The San Ramon total is 30 p-m/capita annually for all buses, and 183 p-m/capita
annually for all transit.
4.4 Transit Leverage
Comparing the VMT for San Francisco to that of San Ramon shows that San Franciscans
saved 5812 VMT per capita as a result of their good transit service and higher densities.
Comparing transit ridership, San Franciscans rode 832 minus 183, or 649 more miles
annually than San Ramon residents. Consequently, at San Francisco's quality of transit
service and density, 649 miles on transit replace 5812 VMT, or 1 mile on transit replaces
9.0 miles of auto use compared to San Ramon. That's a transit leverage of 9 to 1 in
But what is transit's potential to reduce driving in newer suburban areas where
commutes and shopping trips are longer? Comparing Walnut Creek with San Ramon shows that
687 annual p-m on transit reduce driving by 973 VMT per capita, giving a rail suburb to
suburb transit leverage of 1.42 after 17 years of BART to Walnut Creek.
A Walnut Creek to Danville-San Ramon transit leverage of 4 was observed in the 1991
study. Why the decrease this time? It is due to differences in measurements of two
factors: 1) the California Bureau of Automotive Repair measure of VMT/car for Walnut Creek
is much higher in the newer study, 12,175 compared to 10,412; and 2) BART's measure of
Walnut Creek transit ridership is much higher in the new study, 925 compared to 600. Both
of these changes lower the calculated transit leverage. Despite the differences in
calculated transit leverage between the two studies, it is evident that a sizeable
leverage exists even after only 17 years of BART service. The actual Walnut Creek to San
Ramon (or Danville-San Ramon) transit leverage is likely to be between 1.4 and 4.
5. MTC Study Of RAFT Alternative
While developing its Regional Transportation Plan, the Metropolitan Transportation
Commission (MTC) in the San Francisco area analyzed an alternative transportation plan
proposed by the Regional Alliance For Transit (RAFT). By comparing these two plans, and
the projected driving and transit use, a transit leverage can be calculated.
Compared to MTC's alternative, the RAFT alternative would save 200 square miles from
development by clustering the development after 1995 around transit stations. It would
decrease highway development by nearly 500 lane-miles, and build substantially more
transit. RAFT assumed parking cash-out, whereby non-driving employees receive the cash
value of their unused "free" parking space.
MTC's modeling system is one of the most advanced in the country. MTC assesses mobility
and impacts from personal travel with their MTCFCAST model.9
Inputs to MTCFCAST include: ABAG's land use/economic/demographic forecasts;
transportation pricing assumptions; transportation system network assumptions; and travel
behavior assumptions. MTCFCAST is a set of computerized forecasting programs that
simulates travel for an average weekday of a given year. The MTCFCAST system uses
econometric and land use relationships in logit-type models, estimated from their 1982
travel survey. It predicts trips produced, the distribution of trips from point of
production to point of attraction, the mode of travel for each trip, and specific route of
travel between origins and destinations.
MTC projects that the RAFT alternative would reduce daily VMT by 8,829,194 while
increasing daily transit trips by 335,09310.
Assuming that the average transit trip is 6 miles11, the effective transit
leverage of the RAFT alternative compared to MTC's alternative is 4.4 for this 20 year
The results of all these studies are shown in Table 2. They suggest a potential for
strong transit leverage effects, wherein one additional passenger-mile on transit reduces
driving substantially greater than one mile. In every case the transit leverage is higher
than 1, usually much higher. There is scatter in the transit leverage measurements, but
all are in the expected direction. Further, larger transit leverages accompany greater
densification in land uses, and with longer operation of transit service. These confirm
the robustness of the premise. The scatter in the data suggests the need for additional
The studies indicate that VMT reductions of 1.4 to 4 for each p-m on transit can be
achieved within 20 years. Even greater reductions in VMT could accompany higher increases
in mixed-use infill near transit stations. The studies further indicate that transit
leverages ranging from 2.9 to at least 9 can be achieved over the long run.
These results are consistent with the hypothesis that transit leverage results from
land use changes accompanying transit improvements. These land use changes allow shorter
trips, and more pleasant, safe and interesting walks. Such areas have: 1) higher density
to increase the number of nearby destinations; 2) commercial activities mixed with or near
residential neighborhoods, as in traditional towns, to increase the number of nearby
destinations; 3) a pedestrian-friendly neighborhood design with wide sidewalks, weather
and traffic protection for pedestrians, a completed walkway grid offering alternative
routes, and stores close to the sidewalk rather than being set back behind parking lots;
and 4) good transit service. Studies show that residents of such areas drive 1/4 to 1/2 as
much as residents of suburban areas.2,3,4
Reforms are needed to achieve higher density, mixed uses, and pedestrian- and
transit-friendliness. Zoning should allow higher density infill development with local
commerce in or near residential neighborhoods. Mortgage policies should be consistent with
More studies of transit leverage are needed to improve measures of the effects and
better identify the conditions necessary to achieve higher transit leverages to increase
1990 Land Uses, Driving and Transit Ridership
|Pop Density/pop acre
|Net HH density/hh res ac
|BART P-M Capita
|Other P-M Capita
|Total P-M Capita
Population, households, income, autos and total land area are from the 1990 U.S.
census. Residential land area is from Councils of Governments, and excludes streets.
Household density excludes vacant residential units. Transit accessibility is the number
of buses per hour within a quarter mile walk of the average household, or railcars or
ferries within a half mile of the average household. Neighborhood shopping is the fraction
of the community's households which has a mix of five markets, restaurants and/or
drugstores within 1/4 mile walking distance. Pedestrian accessibility includes:
completeness of the pedestrian street grid, sidewalks, hilliness, convenient building
entries, and traffic safety. VMT is annual vehicle miles traveled. BART P-M is the
passenger miles on BART. Other P-M is the passenger miles on other rail transit or buses.
Total P-M is BART P-M plus Other P-M.
Transit Leverage: Reduction in VMT Due to 1 Passenger-Mile On Transit
Transit Leverage (VMT red/p-m on tran)
||Max. Age of Rail System
||NY, Ch, Phil, SF, Bos,Clev
||mixed, up to 104 yr (El)
||urban > 2 mil
||Bos, Chi, NY, SF, DC
||mixed, up to 114 yr
||Den, Det, Hou, LA, Phoe
||23 US, Can, Austr and Euro cities
||mixed, up to 120 yr
||Den, Det, Hou, LA, Phoe
|115 yr (cable)
15 yr (BART)
|117 yr (cable)
17 yr (BART)
|MTC/RAFT 2010 project
||S F region(RAFT assum)
||up to 135 yr(cable car)
||S F region (MTCassum-less rail)
1. Pushkarev, B., and J. Zupan. Urban Rail in America: An Exploration of Criteria for
Fixed-Guideway Transit. Washington: Urban Mass Transportation Administration, U. S.
Department of Transportation, Report No. UMTA-NY-06-0061-80-1, 1980, p 31.
2. Holtzclaw, J. Explaining Urban Density And Transit Impacts On Auto Use. San
Francisco: Natural Resources Defense Council (California Energy Commission Docket No.
89-CR-90), January 1991, p 23.
3. Newman, P. and J. Kenworthy. Cities and Automobile Dependence: An International
Sourcebook. Aldershot, England: Gower Publishing, 1989. Table 3.1, and Data Tables Parts I
4. Holtzclaw, J. Using Residential Patterns and Transit To Decrease Auto Dependence and
Costs. San Francisco CA: Natural Resources Defense Council, June 1994.
5. San Francisco Municipal Railway. San Francisco Municipal Railway Short Range Transit
Plan and Capitol Improvement Program: 1991-2000. September 1991, Tables A-1 to A-3
6. Spiekerman, T., Bay Area Rapid Transit District Planning Department. Letters to John
Holtzclaw, 6 June 1994 and 16 September 1994.
7. League of Women Voters of the Bay Area. "BART Around the Region," Bay Area
Monitor. Lafayette CA: LWV, September 1994, p.6.
8. Howath, C. Contra Costa County Transit Authority. Letter to John Holtzclaw, 7 July
9. Roberts, M.J. 1994 Regional Transportation Plan: Draft Environmental Impact Report.
Oakland CA: Metropolitan Transportation Commission (State Clearinghouse No. 93121113),
April 1994; Metropolitan Transportation Commission. Regional Travel Forecasting Model
System MTCFCAST-80/81: Technical Summary. Oakland CA: MTC, 1988.
10.Dahms, L. "Memorandum to Work Program Committee: Regional Alliance for Transit
Proposal for RTP Track 1." Oakland CA: Metropolitan Transportation Commission, 13 May
11.Purvis, C., Metropolitan Transportation Commission Senior Transportation Planner.
Phone call to John Holtzclaw, 30 August 1994.
Up to Top | Printer-friendly version of this page