REPORT | May 2020

THE ECONOMICS OF URBAN LIGHT RAIL:

A Guide for Planners and Citizens
Connor Harris
Policy Analyst
 The Economics of Urban Light Rail: A Guide for Planners and Citizens

About the Author

Connor Harris is a policy analyst at the Manhattan Institute, where he focuses on infrastructure,
transportation, and housing policy. His writing has appeared in publications such as City Journal,
the New York Post, and the Harvard Political Review. Harris graduated from Harvard University in
2016 with a B.A. in mathematics and physics.

2
 Contents
Executive Summary...................................................................4
I. Introduction ...........................................................................5
II. What Drives Transit Ridership? ..............................................6
III. Common Problems in Light Rail Planning...............................8
IV. Conclusions........................................................................11
Endnotes.................................................................................14

3
 The Economics of Urban Light Rail: A Guide for Planners and Citizens

Executive Summary
Most U.S. cities dismantled their streetcar networks in the first half of the previous century, but in the past few
decades, some transit agencies have restored them, with help from generous federal grants. Modern light rail
systems have appeared in more than a dozen U.S. cities, many of which had little preexisting transit ridership.
They run small trains, often in the middle of wide streets along disused rail lines. Proponents of light rail argue
that it is cheaper than a full subway line and can carry more riders than buses, as well as attracting new riders by
offering a more comfortable ride. One study from 1989, for example, finds that rail systems get 34%–43% better
ridership1 than bus lines with equivalent service.

This trend has had critics, including general opponents of public transit2 and pro-transit urban planners,3 who
argue that light rail service is costlier and less flexible than bus service. This paper aims to help planners and
citizens evaluate when new light rail lines are justified. First, we will explore the economics behind the choice
between light rail and buses, as well as how to determine the best locations of lines. Our analysis of ridership on
23 U.S. light rail systems shows that proximity to development and jobs is necessary for success. Unfortunately,
very few places that are not already served by light rail systems have the features necessary to make them viable
and would be better served by expanded bus service.

Further, we highlight several common errors that light rail planners have made: overexpansion of systems into
low-density areas; overvaluing certain classes of destinations such as airports; and the use of low-quality rights-
of-way that sacrifice ridership to avoid political controversies during construction. These errors have led to the
construction of many existing light rail systems in fundamentally unviable areas.

Although we close by offering a brief list of suggestions for optimizing the efficiency of existing light rail systems
and for evaluating expansions, our analysis ultimately suggests that many proposed expansions, and even some
existing systems, should be canceled or closed. Outside a handful of areas in the U.S., bus rapid transit (BRT)
offers a better, more economically sustainable path forward.

4
THE ECONOMICS OF URBAN LIGHT RAIL:
A Guide for Planners and Citizens

I. Introduction
The abstract economic case for light rail is simple. Though light rail requires more expensive infrastructure than
a bus line, it requires fewer employees to carry the same number of passengers. The Kinki Sharyo light rail trains
in use in Dallas, for example, can hold 94 passengers seated or 274 at maximum crowding, or “crush load,”4 more
than twice what an articulated bus can. Moreover, light rail vehicles, unlike buses, can couple to form longer
trains. In regions where driver wages are expensive relative to infrastructure costs, light rail may be cheaper than
buses by allowing the same capacity with fewer vehicles.

Empirical data, however, call this case into question. One study using an abstract model of a transit network and
cost parameters taken from Sydney, Australia, found that the operational costs of light rail systems, including
both operating costs and the value of passengers’ time, are a few percentage points higher than those of BRT
systems under a wide range of modeling assumptions, except possibly in systems that received more than 3
million passengers per day.5 Another literature review finds an operating cost premium of 59% for light rail over
BRT per unit of capacity, based on studies from 1993 and 2001.6 Data from the 2018 National Transit Database
agree that light rail is more expensive to operate, costing 12%–505% more per vehicle-hour in the cities that were
surveyed. The highest light rail premium is in Oceanside, California, while the median light rail cost premium is
108%, in Pittsburgh.7

Capital costs, likewise, are generally lower for buses than for light rail. One of the least expensive recent light rail
projects is a 14.5-kilometer line in Besançon, France, which opened in 2014. The project cost €249 million, or
roughly $35 million per mile.8 The city government lowered the project cost by soliciting more bids than usual,
using simple, standardized station designs, and minimizing street beautification and reconstruction.9 Typical
light rail projects cost more, though costs vary widely. The recent T7 line in Paris, for example, cost €228 million
for 11.2 kilometers, or about $40 million per mile,10 and the transportation writer Alon Levy gives a range of costs
at about $25 million to $80 million per mile,11 though a planned extension of the T7 is estimated at €223 million
(at 2011 prices) for 3.7 kilometers and six stations, or about $115 million per mile.12

Costs in the U.S. are generally somewhat higher. The Valley Metro Rail project in Phoenix, completed in 2008,
cost about $85 million per mile in today’s money. The final stage of the DART Orange Line extension to Dallas–
Fort Worth (DFW) airport, which was finished in 2014, cost $83 million per mile in today’s money, though this
project included only one station and would have cost more with a more typical station spacing.13

Because buses rely on preexisting roads, the cost of a bus line, at minimum, is just the cost of buses themselves
as well as bus shelters. Nevertheless, many cities have implemented BRT systems, which include infrastructure
improvements such as fancier stations, off-board fare-collection machines, raised boarding platforms for mobil-
ity-impaired passengers, and dedicated lanes.

5
 The Economics of Urban Light Rail: A Guide for Planners and Citizens

These systems can cost a large fraction of what light to pass each other—but installing such systems in most
rail does, though they are still generally cheaper. Costs downtown areas is impossible or disruptive. A 2009
of BRT projects vary widely: one paper from 2005, for literature review suggests that, overall, light rail has a
example, found that costs varied by almost an order of 65% capacity increase over BRT.18
magnitude.14 A 2001 report from the U.S. Government
Accountability Office gave an average cost of BRT Despite high potential capacity, actual ridership on
with full busways as $13.5 million per mile (about $20 U.S. light rail systems is low. Boston’s Green Line—
million today), about 39% of the $34.8 million cost which is a small component of a far more extensive
of light rail (about $50 million today).15 More recent and faster subway system—had 6,951 boardings per
systems have tended to be more expensive: a survey route-mile on an average weekday in 2018. Only four
of recent U.S. BRT systems by the planner Ari Ofsevit other systems garner more than 3,000: San Fran-
found that costs varied from $27 to $60 million per cisco (5,088), Seattle (3,768), Minneapolis (3,411),
mile.16 There was one low-end outlier, at $6 million and northern New Jersey (3,032, including the Hud-
per mile: a narrow one-lane busway in Eugene, son-Bergen Light Rail and the Newark Subway).19
Oregon, whose planners scheduled buses carefully so
that buses could run in both directions despite only Most trams in Europe, even those in very small cities,
having one lane. perform much better. In France, for example, the
trams in the Paris region, located in relatively periph-
Nevertheless, in general, the difference between light eral areas, garnered about 12,600 riders per route-mile
rail and BRT capital costs is about $30 million per on an average weekday in 2017.20 The 41.1-mile Bor-
mile, with a wide margin of variability. This is an ad- deaux tramway in Bordaeaux, meanwhile, received
ditional $1.37 in capital costs per ride if amortized 105.5 million riders in 2018, or an average of 7,000
over 20 years with daily ridership of 3,000 riders per per day per route-mile, including weekends.21 Even
route-mile, typical for the higher-performing U.S. most light rail systems in Canada get much higher rid-
light rail lines. ership than in the United States. Calgary’s CTrain, for
example, is 36.7 miles long and got 312,300 riders on
These conclusions should be taken with a grain of an average weekday in the fourth quarter of 2018, or
salt. Construction costs vary widely, and, in certain about 8,500 per mile.22 Edmonton’s Light Rail Transit,
circumstances (for example, if a preexisting rail line meanwhile, averaged 110,786 riders per weekday in
could be converted easily for light rail), light rail may 2018, or about 7,337 per mile.23 These figures call into
be cheaper. But on the whole, transit agencies should question whether most U.S. cities have any transit cor-
assume that replacement of bus service with light rail, ridors busy enough to merit conversion from buses to
unless buses are already overcrowded, will increase op- light rail.
erational costs.

Improved capacity and reliability may still make

the conversion worthwhile. A phenomenon called II. What Drives Transit
“bunching” limits the frequency of bus service. If a
bus is slightly delayed—say, by missing a traffic light, Ridership?
or by a disabled passenger who needs additional
time to board—more passengers will accumulate at There are two principal drivers of transit ridership.
subsequent stops while the bus is delayed. The bus First, and most important, is development density sur-
must then stop longer to let the extra passengers board, rounding the stations. When travelers have to walk or
causing a vicious cycle of delays and overcrowding. drive long distances to reach a transit station, driving is
Eventually, less crowded buses farther back will catch almost always faster, even in congested traffic. This is
up to the delayed bus in front. especially true of commuters, who are reluctant to have
to travel long distances between their jobs and down-
Some BRT improvements can ameliorate minor bunch- town transit stops (they are generally more tolerant of
ing by providing more reliable travel and boarding long trips from transit stops to their homes). This ob-
times. Nevertheless, bunching remains a fundamen- servation is commonly called the “Quarter-Mile Rule”:
tal cap on bus capacity. One expert on BRT suggests most commuters on public transit travel no more than
that bunching makes the frequency of more than 30 a quarter-mile from a transit stop to their jobs, and few
buses per hour impossible, and 22 is ideal.17 Some BRT travel more than 700 meters, or about 0.43 miles.24
systems manage higher frequencies by using freeway
lanes with no traffic crossings, closing most cross Second is frequency of service. For short-distance trips
streets, or by using four lanes of a road to allow buses common on U.S. light rail systems—in 18 of the 23 U.S.

6
systems studied here, the mean trip length is under ridership may not be reliably separable with simple sta-
seven miles—time spent waiting for vehicles can far tistical tools. Our model, however, estimates that for a
outweigh time spent traveling. More frequent service, typical system with 2,000 riders per day per route-mile
then, has a greater effect on overall trip time than faster and 100 daily trains, an increase to 110 daily trains
speeds. (Waiting-time decreases have a still larger would improve ridership by approximately 6%, for an
effect on riders’ perceived travel time: studies con- elasticity of 0.6. This is within the range of elasticities
sistently find that riders perceive a minute of waiting found by other studies.
time as longer than a minute of in-vehicle time. Some
studies find that riders value waiting time more than
twice as much as in-vehicle time.)25 Why does the U.S. do so poorly?
Researchers at McGill University, for example, com- U.S. light rail systems perform very poorly relative
pared frequency and ridership statistics for several to most European systems, but also compared with
North American bus systems and found: “The volume systems in Canada that serve cities similar to U.S.
of service delivered by a transit agency in a region is cities. Edmonton and Calgary, for example, resemble
by far the largest factor explaining public transit rid- U.S. cities in their high rates of car ownership and pre-
ership in the North American metropolitan regions we dominance of single-family houses. The population
studied.”26 Other studies have investigated the elasticity density of Calgary is 3,442 residents per square mile,
of transit ridership relative to service changes. (“Elas- and that of Edmonton is 3,525 residents per square
ticity” is the percentage increase in ridership caused by mile, both comparable with Phoenix, which had 3,207
a 1% increase in service frequency.) One experimental residents per square mile within the city boundaries in
study in Minneapolis estimated an elasticity of about 2018. Despite this, both Edmonton and Calgary have
0.39;27 that is, 1% more service means 0.39% more higher rates of light rail ridership.
riders. If the marginal costs of an additional train are
less than 39% of the average costs—quite possible, Different urban-planning decisions may help explain
as fixed costs such as track maintenance account for the disparity. U.S. planners devoted vast resources
large portions of rail costs—running additional service toward facilitating car access to downtowns, build-
would be net profitable. Other reviews have found even ing massive expressways, and leveling city blocks for
higher elasticities; one study, for instance, estimates a more parking space. The zoning codes in all U.S. cities
long-run elasticity of 0.7–1.1.28 also require commercial and residential developers to
provide large amounts of off-street parking, often far
These observations are borne out by a simple linear in excess of market demand. Parking minimums effec-
model of light rail ridership, using data from 23 U.S. tively subsidize driving at property owners’ expense
light rail systems, as shown in the figure on the next and reduce feasible development densities, thereby
page. The graph suggests that service frequency and suppressing transit demand.
development of density around light rail stations ac-
counts for most of the variation in transit ridership, and In Edmonton and Calgary, by contrast, planners kept
shows why light rail construction in low-density areas expressways out of downtown. Calgary also deliberate-
is typically unsuccessful. It is possible to predict 89% ly restricted parking construction in the downtown area
of the variance in their performance almost perfectly and encouraged the redevelopment of surface parking
from three variables: the total number of jobs within lots as buildings, thereby keeping parking prices high.29
a half-mile of each station; the number of employed Though Canadian zoning codes do include parking
workers within a half-mile; and the total number of requirements and density limits, they tend to be less
trains per day per direction (calculated by dividing the onerous than in the United States. For example, the
total number of train-miles run in an average day by code that governs most single-family areas in Calgary
twice the route length of the system). requires only 330 square meters of land per house,
or about 12 houses per acre, much higher than the
These figures should not be overinterpreted. Using standard densities of suburban developments in the
radii different from a half-mile gives equally good United States. In California, by contrast, the median
predictions. minimum lot size is 6,000 square feet, allowing about
seven houses per acre.30
Furthermore, frequency of service itself is correlat-
ed with other variables that predict ridership: denser
areas will naturally merit higher levels of service, and
the effects of denser development and better service on

7
 The Economics of Urban Light Rail: A Guide for Planners and Citizens

Predicted and Actual 2017 Weekday Ridership per Mile

SF SEA: Seattle, WA
MSP: Minneapolis–Saint Paul, MN
5,000 NJ: New Jersey
HOU: Houston, TX
LA: Los Angeles
PHL: Philadelphia
4,000 BUF: Buffalo, NY
Actual Weekday Ridership

PDX: Portland, OR
MSP SEA SD: San Diego, CA
CLT: Charlotte, NC
3,000 NJ
DEN: Denver, CO
HOU
SLC: Salt Lake City, UT
LA
BUF PIT: Pittsburgh, PA
SD PDX PHL STL: St. Louis, MO
2,000
SJ: San Jose, CA
DAL: Dallas, TX
CLT BAL: Baltimore, MD
SLC
PIT DEN
1,000 DAL SJ: San Jose, CA
SAC
STL NOR: Norfolk, VA
BAL SJ
SAC: Sacramento, CA
NOR CLE: Cleveland, OH
OCE CLE
OCE, Oceanside, CA
0 2,000 4,000
Predicted weekday ridership from service frequency and development density

Source: Author’s Calculations based on U.S. Census Bureau, Longitudinal Employer-Household Dynamics and American Public
Transportation Association, “Public Transportation Ridership Report” Fourth Quarter, 2018.

III. Common Problems Dallas’s DART system (1,024 weekday riders per route-
in Light Rail Planning mile in 2018) and Denver’s RTD (1,443 weekday riders
per route-mile) are two good examples. Christof
Spieler, a planner for Houston’s public transit agency,
Light rail works best in a very particular set of has detailed the political pressures that led both
circumstances. In areas where driving is difficult but systems to expand too far. In DART’s case, the system
travel demand is large and consistent, it can facilitate was financed by each participating municipality and
relatively short trips of a few miles or less. In these had an incentive to cover as many municipalities as
cases, the ability to run larger light rail vehicles with possible—often in undesirable locations on corridors
less bunching and fewer drivers can offset the increased cobbled together out of old freight rail rights-of-
expenses in capital maintenance. Over longer distances, way.31 Many leading Dallas politicians, meanwhile,
the low speed of light rail vehicles cannot compete with had invested in real-estate development in Dallas’s
freeways; with lighter traffic loads, the problem of bus northern suburbs, so they had incentives to push
bunching becomes less severe. for an expansive, suburb-focused rail network.32 The
managers of RTD in Denver, meanwhile, explicitly
The low ridership of most U.S. light rail systems, conceived of light rail as a commuter rail substitute,
however, frequently stems from one of two mistakes: and its lines reach far out into low-density residential
overexpansion; and route choices that are more neighborhoods. As a result, most Denver stations have
politically expedient than economically efficient. low ridership: of the 66 stations in Denver’s frequent
rail network, only 16 average more than 500 boardings
per weekday. This is even worse than DART, where 55
Overexpansion of 64 stations get more than 500 riders per day.33

The best-performing newer systems in our database, Two species of overexpansion in U.S. cities deserve
such as Minneapolis, Seattle, and Houston, are all special mention. First is an overemphasis on serving
compact, serving urban areas near downtown. By transit-oriented developments. Many cities have seen
contrast, larger light rail systems that stretch into low- new developments on “New Urbanist” principles:
density suburban areas tend to underperform. apartments with mid-rise units and a mix of commer-

8
9
 The Economics of Urban Light Rail: A Guide for Planners and Citizens

cial and residential development aimed at satisfying industrial suppliers and distribution centers, demand
most residents’ daily needs without having to drive. large amounts of land and are difficult to access on
Many of these developments are also transit-oriented, foot, making them low-value destinations for transit
to allow for travel outside the development, such as to ridership. Finally, airport noise and pollution make
downtown jobs. the surrounding areas less desirable for the sort of
redevelopment that might improve ridership.36
Because of strict zoning laws in developed areas
of cities, these developments often must be built DART, again, is an instructive example. In 2014,
miles from established downtowns. As such, transit- the system unveiled an approximately 14.5-mile
oriented developments frequently disappoint. New, expansion to DFW airport. But the airport station
isolated developments are rarely large enough to has middling ridership: in fiscal year 2018, it saw 976
be self-contained or offer the amenities of true city boardings per average weekday, the 34th of DART’s
centers. Residents who want to travel to specialty 64 stations, and less than half the ridership of
stores or jobs not readily accessible by the existing stations in some moderately dense residential areas.
transit network—and in typical low-density U.S. The route to the airport, furthermore, runs through
cities, this is almost all of them—will need to own cars. sparsely populated suburbs and along a freeway.
Once they own cars, there’s no reason not to use them Some stations have barely any development within
for all trips, especially if zoning policies guarantee a quarter-mile. Three of the five non-airport stations
copious parking. on the line rank in the four lowest-ridership stations
in the entire system. Altogether, the extension
For example, DART’s Las Colinas Metro Center receives only 2,510 riders per average weekday (for
station serves a transit-oriented development in an comparison, a single full train in the NYC subway
isolated area about 12 miles from downtown Dallas. has a capacity of about 2,000 riders). It costs about
The trip from Las Colinas to downtown takes about $1.8 billion, or $700,000 per average weekday rider
35 minutes, almost always slower than driving, even (for comparison, tram lines in Paris typically cost
in rush hour. The station is surrounded by mid- $10,000 or less, per rider).37
rise apartment buildings and office towers, but its
ridership is the second-lowest in the system. Recent Denver and Salt Lake City are also instructive. The
extensive construction around the station has not had 23.5-mile-long “A Line” of Denver’s RTD system—
a large impact on ridership, which increased from 236 strictly speaking, a commuter rail rather than a light
boardings per average weekday in 2014 to only 266 rail system, as it runs large trains with a higher top
in 2018.34 Transit-oriented development around the speed—serves the airport, located in the middle
Cityline station in the Dallas suburb of Richardson, of undeveloped prairie 18 miles from downtown
about 16 miles from downtown, coincided with a Denver. The airport station itself is scarcely better
decline in ridership.35 The worst-performing station used than the station at DFW: between May
in Denver’s RTD network is Lone Tree Station, which and August 2019, it received an average of 1,819
stands in an empty field, having been constructed to boardings per day, Monday through Thursday.
serve an incomplete transit-oriented development on (Nevertheless, the airport station is the second most
the southern fringe of the urban area. used in the whole system, showing how poorly the
system is performing overall.) The airport station is
Another form of overexpansion comes from the also far-flung: between it and the next station on the
tendency of light rail planners to overvalue airport line, which serves an exurban housing development,
service. It’s easier to form a broad political coalition are almost eight miles of nonstop track. The A
for airport service than for run-of-the-mill transit Line runs on a repurposed preexisting freight line
improvements. City power brokers like to impress flanked by industrial zones, which are forbidding for
out-of-town visitors with airports, and suburban pedestrians. In total, the line draws fewer than 1,000
residents who do not use transit regularly imagine riders per mile.
that a train for their occasional airport trips would be
convenient. Salt Lake City’s TRAX system is similar: one of its
branches connects the city’s airport to downtown.
Airport connectors, however, tend to perform poorly. The airport is relatively close to downtown, but the
Airports are usually in remote locations, so light rail journey is still about 5.5 miles, including a 3.3-mile
to airports requires extensive capital construction; the nonstop segment. The extension receives middling
slow speeds of light rail relative to freeways matter more ridership, even by the standards of a light rail system
for long-distance trips from airports to downtown. with generally poor performance. The airport station
Moreover, businesses that surround airports, such as averaged 1,355 daily boardings in 2018, ranking 16th

10
in the system out of 58 total stations. The other four runs through the southern suburbs of L.A.; two of its
stations on the line ranked 25th, 27th, 37th, and stations are located inside freeway interchanges. In
50th, with the poorest-performing station receiving Portland, about 14 miles of the MAX light rail system,
only 168 riders on an average day in 2018. including the main trunk that carries three lines to the
city’s eastern suburbs, are built either in the median
or alongside Interstates 84 and 205. Much of Denver’s
Avoiding dense, walkable areas for light rail system, one of the nation’s worst-performing,
and a significant underperformer relative to surround-
ease of construction ing development density, runs along Interstates 25 and
225.
The best place to build light rail, as our findings on the
association between density and ridership show, is in Light rail lines along freeways are undesirable for
dense locations near jobs. Unfortunately, this inevi- several reasons. First, the freeway takes up much of
tably involves taking street space away from cars and the land accessible from rail stations on foot. Second,
thus is often controversial. The temptation always because freeways are convenient to access by automo-
exists for planners to use easier alignments instead, bile but unpleasant to live near, they tend to be sur-
such as building along freeways or abandoned freight rounded by lower-value land uses. Third, trips on a
railways, or through industrial zones with few resi- light rail line that runs alongside a freeway are compet-
dents to object. ing directly with the region’s fastest car trips. The light
rail systems in Seattle, Minneapolis, Houston, and San
Charlotte’s LYNX system, one of the lowest- Francisco, which are all among the best-performing in
performing of the 23 systems we examined, shows how the United States, have no segments, or only short pe-
this temptation manifests. South of downtown, much ripheral segments, along freeways.
of the city’s light rail line runs alongside an active
Norfolk Southern freight line, roughly parallel to The undesirability of lines located in freeway medians
South Boulevard, an important commercial street. But can be seen in Los Angeles’s Green Line, which has
accessing South Boulevard from the light rail stations the weakest ridership in the system: it averaged just
requires walking several hundred feet or more, and the 1,542 riders per weekday per route-mile in 2018. The
freight line sometimes cuts off the easiest pedestrian Gold Line (which also runs some distance in a freeway
routes. Moreover, though many LYNX stations have median), by comparison, had 1,630 riders per weekday
seen considerable redevelopment, the southern end of per route-mile; the Blue and Expo lines, respectively,
the route remains undeveloped. Dallas’s DART Green averaged 2,939 and 4,015 riders per route-mile.39 (The
Line, which runs northwest of the city, likewise uses comparison is admittedly imperfect, as the Green Line
a freight rail corridor through a large industrial zone. is also the only one that does not serve downtown Los
The stations along this line are among the least busy Angeles.)
in the system: of the four stations within the industrial
zone (Downtown Carrollton, Farmers Branch, Royal
Lane, and Walnut Hill/Denton), none ranks above
42nd out of the system’s 64 stations in ridership. IV. Conclusions
Planners also sacrifice destination quality for expedi- There may be a restricted role for light rail in U.S.
ence by running light rail lines along peripheral roads cities: handling frequent trips in dense inner urban
or freeways. Christof Spieler has observed that many areas, where passenger demands exceed bus capaci-
high-value destinations in Denver are served only indi- ty and the higher speed of freeways is less of a selling
rectly. For example, the R light rail line skirts one side point. But most light rail systems do not fit this niche:
of the Anschutz Medical Center, along a broad road, they serve low-ridership corridors, or long-distance
rather than running through it;38 much of the medical trips that cannot compete with freeways.
center is thus over a mile from the station that nom-
inally serves it. The main system hub, Denver Union Most transit agencies should be far more skeptical of
Station, is located on the northwest edge of downtown; building new light rail lines than they have been; any
many riders must transfer to a free shuttle to get to corridor without enough ridership to fill a bus every
their final central-city destinations. few minutes should not be a light rail line. Light rail
projects that do not serve any current ridership should
Some cities have built substantial portions of their light not be defended on the grounds that future transit-ori-
rail networks along freeways. In Los Angeles, the Green ented development will provide a source of riders. Rid-
Line is mostly in the median of Interstate 105, which ership from these developments is often disappointing,

11
 The Economics of Urban Light Rail: A Guide for Planners and Citizens

and many residents of far-flung transit-oriented devel- In Seattle, finally, Sound Transit has proposed extend-
opments would be much happier living in apartment ing the light rail line south from its current terminus at
buildings in inner-ring neighborhoods if zoning poli- SeaTac airport to the suburb of Federal Way, 20 miles
cies were more rational. For aiding commutes from from downtown.43 The extension lacks the virtues that
far-flung suburbs, additional commuter bus service made the existing Seattle light rail network relatively
would be faster and cheaper than light rail expansions. successful. The planned 7.8-mile, $3.16 billion light
rail line—at $400 million per mile, an inordinate con-
It may be better to simply tear down some especially struction cost for an aboveground line—will run along-
low-performing lines run by transit agencies with high side Interstate 5 through low-density residential and
light rail operating costs. The savings can then be redi- retail areas with poor pedestrian accessibility. All three
rected to running more frequent service on core parts cities should reconsider or halt plans for expansion.
of the line or simply to reducing the need for tax sub-
sidies. Such a move would not be wholly unprecedent- Though most transit agencies should not expand light
ed. In a recent redesign of its network, San Jose closed rail systems, many of them could nevertheless improve
down a particularly low-performing outer stretch of its the use of current lines. For example, adding more fre-
light rail line.40 Other agencies with extremely far-flung quent off-peak service is very cheap.44 Infrastructure
lines that receive little ridership, such as DART’s line to and rolling stock expenses are largely determined by
DFW airport, may want to consider closing these exten- the system’s peak capacity, as are crew expenses, since
sions and instead serving the more important stations hiring split-shift train drivers, who work only in rush
with express shuttle buses. Some smaller systems may hours and break during midday, is difficult. Neverthe-
be better off closed entirely. For example, the Sprinter less, many systems run very low frequencies during
light rail line in northern San Diego County is the low- off-peak hours. DART, for example, runs only once
est-performing light rail system in ridership per mile of every 20 minutes; Phoenix’s Valley Metro Rail runs
the 23 we have studied, and it has the second-highest every 24 minutes. Given the low marginal costs of off-
operational expenses per mile and the highest ratio of peak service as well as the high elasticity of ridership
light rail to bus expenses. Sprinter also parallels a sig- to service frequency, higher off-peak frequencies on
nificantly faster road—the Ronald Packard Parkway— many systems would quite likely pay for themselves in
and the surrounding neighborhood has shown little improved ridership.
appetite for high-density redevelopment. Officials in
the region should strongly consider replacing Sprinter Finally, cities with light rail systems should eliminate
service with bus service on parallel roads and exper- all regulatory barriers to using the land around them
imenting with other methods of congestion control, to the fullest. This means allowing high-density con-
such as dynamic tolling. struction in regions adjacent to transit stops and elim-
inating hidden subsidies for driving, such as parking
Unfortunately, some agencies are planning light rail minimums. As UCLA urban-planning professor
expansions to corridors with unproven, and likely low, Donald Shoup has documented, the off-street parking
demand. In Phoenix, for example, voters recently de- minimums common in most U.S. zoning codes severe-
feated a ballot measure that would have halted light ly inflate the costs of new residential and, especially,
rail construction, allowing the expansion of the city’s commercial construction.45 They also handicap light
light rail system to the south and the west. Unfortu- rail by subsidizing driving and often force new devel-
nately, both expansions run through low-value land. opment into low-density, pedestrian-hostile forms,
The southern extension serves a low-density area cur- such as big-box retail with stores surrounded by large
rently filled with vacant lots, and the agency itself proj- parking lots. Houston has taken a step in the right di-
ects that the project will cost $81,000 per daily rider.41 rection by abolishing parking minimums in the Down-
The planned western expansion, meanwhile, runs town, Midtown, and East Midtown neighborhoods, all
mostly alongside Interstate 10 through largely indus- of which are served by light rail.
trial areas.
Improving public transportation is a worthy goal.
In Houston, the cornerstone of a recently approved But far-flung light rail networks through industrial
light rail system is a line to Hobby Airport through zones—though they may get support from suburban
industrial and low-density residential areas, estimat- residents who think that more transportation options
ed to cost $167,000 per daily rider.42 More promising to downtowns and airports are always better—are
corridors, such as the long-proposed Universities Line counterproductive and expensive. The political
that connects several of Houston’s most important job compromises required to get light rail lines built,
centers, are receiving only bus upgrades. furthermore, often result in bad routes that are worse
than no line at all. Other than a few dense downtown

12
and near-downtown areas, most of which already
have light rail lines, nowhere in the nation has
enough public transit demand to justify the cost of rail
investment. Public transit advocates should focus on
more cost-effective methods for serving low-density
suburban areas and on reforming land-use regulations
to optimize the use of existing systems.

13
 The Economics of Urban Light Rail: A Guide for Planners and Citizens

Endnotes
1 Edson L. Tennyson, “Impact on Transit Patronage of Cessation or Inauguration of Rail Service,” Transportation Research Record 1221 (1989).
2 Randal O’Toole, “The Worst of Both: The Rise of High-Cost, Low-Capacity Rail Transit,” Policy Analysis (Cato Institute), no. 750 (June 3, 2014).
3 Jarrett Walker, “Streetcars: An Inconvenient Truth,” Human Transit (blog), July 3, 2009.
4 Dallas Area Rapid Transit (DART), Capital Planning Division, DART Reference Book, Version 10.0, March 2019.
5 Alejandro Tirachini, David A. Hensher, and Sergio R. Jara-Díaz, “Comparing Operator and Users Costs of Light Rail, Heavy Rail and Bus Rapid Transit
over a Radial Public Transport Network,” Research in Transportation Economics 29, no. 1 (2010): 231–42.
6 Ming Zhang, “Bus Versus Rail: Meta-Analysis of Cost Characteristics, Carrying Capacities, and Land-Use Impacts,” Transportation Research
Record 2110, no. 1 (January 2009): 87–95.
7 Federal Transit Administration, “NTD Data: Expenses.”
8 Keith Barrow, “Besançon Inaugurates ‘Low Cost’ Tram Line,” International Railway Journal (Sept. 1, 2014).
9 Yonah Freemark, “How Is Besançon Building a Tramway at €16 million/kilometer?” The Transport Politic (blog), Oct. 1, 2010.
10 Planitec BTP, “Design and Construction of the T7 Tramway Line Between Villejuif and Athis-Mons.”
11 Alon Levy, “What Are the Strong Tramway Corridors?” Pedestrian Observations (blog), July 20, 2016.
12 Régie Autonome des Transports Parisiens (RATP), “T7 Extension to Juvisy-sur-Orge,” Sept. 18, 2018.
13 U.S. Dept. of Transportation (DOT), “Dallas Area Rapid Transit Project Orange Line Extension (I-3).”
14 Daniel B. Hess, Brian D. Taylor, and Allison C. Yoh, “Light Rail Lite or Cost-Effective Improvements to Bus Service? Evaluating Costs of Implementing Bus
Rapid Transit,” Transportation Research Record 1927, no. 1 (2005): 22–30.
15 Ibid.

16 Ari Ofsevit, “How Much Does BRT Cost? 7X Less than LRT? (Hint: No),” The Amateur Planner (blog), June 3, 2015.
17 Walter Hook, “Bus Rapid Transit Planning Guide: 2016 Edition,” BRT Planning International, 2016.
18 Zhang, “Bus Versus Rail.”
19 “Wikipedia: List of North American Light Rail Systems by Ridership,” Wikimedia Foundation, last modified Mar. 4, 2020. All numbers are for 4Q:2018.
20 Olivier Razemon, “En région parisienne, le tramway victime de son succès,” Le Monde, June 19, 2019.
21 Denis Lherm, “Tramways et bus à Bordeaux : 168 millions de voyageurs en 2018, le trafic explose,” Sud Ouest, Mar. 28, 2019.
22 American Public Transportation Association, “Public Transportation Ridership Report” Fourth Quarter, 2018.
23 “2018 LRT Passenger Count Report,” Urban Form and Corporate Strategic Development, City of Edmonton, March 2019.
24 Reinhard Clever, “The Last Mile Falling Through the Cracks—Case Study of the San Francisco to San Jose Section of the California High Speed Rail
System,” presentation at the 90th annual meeting of the Transportation Research Board, Washington, D.C., Jan. 24, 2011.
25 Yingling Fan, Andrew Guthrie, and David Levinson, “Waiting Time Perceptions at Transit Stops and Stations: Effects of Basic Amenities, Gender, and
Security,” Transportation Research Part A: Policy and Practice 88 (2016): 251–64.
26 TransitCenter, “Service Matters: Examining Ridership Trends in 25 North American Cities over Time,” May 29, 2018.
27 Joseph Totten and David Levinson, “Cross-Elasticities in Frequencies and Ridership for Urban Local Routes,” Journal of Public Transportation 19, no. 3
(2016): 117–25.
28 Todd Litman, “Transit Price Elasticities and Cross-Elasticities,” Victoria Transport Policy Institute, Mar. 18, 2019, p. 20.
29 Neil McKendrick et al., “Calgary’s CTrain—Effective Capital Utilization,” Calgary Transit, 2007.
30 Salim Furth and Olivia Gonzalez, “California Zoning: Housing Construction and a New Ranking of Local Land Use Regulation,” Mercatus Research,
Mercatus Center at George Mason University, August 2019.
31 Quoted in Matt Goodman, “DART Successfully Built for Sprawl. So When Will It Design for People?,” D Magazine, Nov. 30, 2018.
32 Jim Schutze, “Dallas Needs to Do a Tammy Wynette on DART—D-I-V-O-R-C-E,” Dallas Observer, Feb. 13, 2020.
33 Regional Transportation District Denver, “Rail Station Activity Analyzed,” May 2019.
34 DART Capital Planning Division, DART Reference Book, Version 7.0, March 2016; DART Reference Book, Version 10.0.
35 Christof Spieler, Trains, Buses, People: An Opinionated Atlas of U.S. Transit (Washington, D.C.: Island Press, 2018), 105.
36 See, e.g., Ermanno Affuso et al., “Is Airport Proximity an Amenity or Disamenity? An Empirical Investigation Based on House Prices,” Land
Economics 95, no. 3 (2019): 391–408; Jacek Batóg et al., “Investigating the Impact of Airport Noise and Land-Use Restrictions on House Prices:
Evidence from Selected Regional Airports in Poland,” Sustainability 11, no. 2 (2019): 412.
37 Alon Levy, “Why It’s So Expensive to Build Urban Rail in the U.S.,” CityLab, Jan. 26, 2018.
38 Quoted in Andy Bosselman, “Denver’s Buses and Trains Are Not Useful to Most People. A New Book Shows Why,” StreetsBlog Denver, Jan. 29, 2019.
39 Los Angeles County Metropolitan Transportation Authority, “Interactive Estimated Ridership Stats.”
40 Jarrett Walker, “San Jose and Silicon Valley: Welcome to Your New Network,” Human Transit (blog), Dec. 28, 2019.
41 U.S. DOT, “South Central Light Rail Extension: Phoenix, Arizona.”
42 Connor Harris, “Metro’s Planned Light Rail to Hobby Airport Is a Bad Investment,” City Journal Online, June 30, 2019.
43 U.S. DOT, “Federal Way Link Extension: Seattle, Washington.”
44 Alon Levy, “Base Train Service Is Cheap, Peak Train Service Is Expensive,” Pedestrian Observations (blog), Jan. 22, 2018.
45 Donald Shoup, “Parking Reform Will Save the City,” CityLab, Sep. 20, 2019.

14
15
May 2020