The Antiplanner

Dedicated to the sunset of government planning

Antiplanner Policy Brief Number 111 July 27, 2021

Cost Overruns and Ridership Shortfalls

R ail transit projects built in the United States typical-
ly suffer severe cost overruns and end up carrying far
fewer riders than originally projected. The latest studies
Six years ago, the Antiplanner presented a summary of
these reports on capital cost overruns and ridership short-
falls. Since then, the FTA has issued reports on 13 more
published by the Federal Transit Administration (FTA) in- projects, the latest of which was written in 2020 for a proj-
dicate that the projections made for some recent projects ect completed in 2015. All 75 projects reviewed by Pickrell
are better than those made in the past. However, this is or the FTA are shown in a table on page 3.
partly because the FTA has changed its definition of “cost Transit Project Optimism Bias, 1986-2015
overrun” and partly because the FTA has not yet looked at 60%
some projects that we know have huge overruns, such as 40%
the Honolulu rail project.
20%
The Department of Transportation first looked at
this issue in a 1990 report by Don Pickrell, who looked at 0%
four heavy rail, four light rail, and two automated guide- -20%
way (“people mover”) projects in nine cities. On average, -40%
Pickrell found, building these projects ended up costing
-60%
62 percent more than projected, operating them cost 130 1980s 1990s 2000s 2010s
percent more than projected, and ridership was 47 percent Cost Overruns Ridership Shortfalls
less than projected.
“The systematic tendency to over-estimate ridership Transit projects completed since 2010 appear to have been based on
and under-estimate capital and operating costs introduc- more realistic estimates than earlier ones. However, this is really due to a
change in the baseline used by the FTA in its analyses.
es a distinct bias toward the selection of capital-intensive
transit improvements such as rail lines,” observed Pickrell. The numbers show that projects completed in the
“Rail becomes the economically preferred transit mode 2000s had cost overruns averaging 38 percent and rid-
only when its substantial capital costs and fixed operat- ership shortfalls of 44 percent. Projects completed in the
ing expenses can be spread over large passenger volumes.” 2010s had cost overruns of only 17 percent and ridership
Thus, even if estimates for bus or other low-cost modes are shortfalls of only 5 percent. However, this apparent im-
just as poorly estimated as for rail, “the planning process provement may be due to a change in how the FTA de-
will still be biased toward selection of the most capital-in- fined cost overruns.
tensive alternatives under consideration.”
Pickrell’s report was so controversial that he was Which Projection Should Be Used?
transferred to another part of the DOT and told never to Transit agencies make several estimates of costs and rid-
write about transit issues again. But debate over rail transit ership over the course of planning and building a project.
forced the FTA to repeat Pickrell’s analysis for more re- Estimates might be made when projects are first proposed,
cent projects in reports issued in 2003, 2008, 2011, 2012, when they are compared with other alternatives, when
and 2013. These included 52 projects completed as late as the draft and final environmental impact statements are
2009. The projects included a handful of bus-rapid transit prepared, when applications for federal grants are made,
lines, but most were some form of rail transit. Not all rail immediately before construction begins, and during the
projects were reviewed, but there is no indication that the construction period.
FTA has deliberately biased its sample to include projects Successive estimates of costs tend to rise while rider-
with smaller cost overruns or ridership shortfalls. ship estimates fall. Some agencies take advantage of this,
claiming they completed a project under budget when cantly underestimating the cost overruns.
what they mean is they completed a project for less than
the cost projected when or after construction began, even Cost Trending Upward
though that cost may be much higher than earlier in the Counting all projects, and after adjusting for inflation,
planning process. construction costs per mile were significantly higher in
The FTA’s earlier analyses looked at the cost projec- the 1980s than the 2010s. But this is because the 1980s
tions made when agencies were comparing alternatives. included several expensive heavy-rail projects while the
Ideally, this step includes a comparison of rail with bus 2010s had no heavy rail but instead included several rela-
and possibly even highway improvements. Once the agen- tively inexpensive streetcar and bus projects.
cy selects rail, the other alternatives are dropped. Although the projects reviewed by the FTA span the
The FTA’s more-recent analyses used the cost projec- better part of four decades, light rail is the only technology
tions made at a step known as “PE-entry,” that is, the be- reviewed in all four decades. Counting only light-rail proj-
ginning of preliminary engineering. By this step, the agen- ects, average costs per mile in the 2010s were $82 million
cy has discarded all other alternative modes, and the only per mile, which was 40 percent more than the 1980s and
alternatives to be considered are different routes. With no almost 90 percent more than in the 1990s.
competition from other alternatives, costs can be higher Inflation-Adjusted Light-Rail Costs
without overtly admitting that bus or some other mode 100
might be better. Thus, one of the reasons why cost over-

Millions of Dollars Per Mile

runs appear to have declined in recent years is that the 80
FTA is using cost projections made at a later stage in the
process. 60
For example, in 1997, Denver’s Regional Transit Dis-
trict (RTD) published alternatives analyses (then known 40
as major investment studies) for rail lines proposed to go
from the Denver airport through downtown Denver to 20
Wheat Ridge, Colorado. These were known as the “East”
and “Gold” lines but eventually were built under one con- 0
1980s 1990s 2000s 2010s
tract. The major investment studies estimated that con-
structing the lines would cost less than $500 million. The average cost of light-rail lines completed after 2000 was much high-
er than earlier lines. This assumes that the lines included in the FTA’s
RTD decided to build the lines and dropped bus and
before-and-after reports are representative of those completed during
highway alternatives, leaving open only the question of these decades.
whether the trains would be powered by Diesels or elec-
tricity. In 2004, RTD asked voters to approve funding for This probably understates the increase in costs over
the lines, by which time RTD projected the lines would this period as seven of the eight 2010s light-rail projects
cost $1,165 billion, which after adjusting for inflation was selected for review by the FTA had unusually low costs per
31 percent more than the major investment study esti- mile. After adjusting for inflation to today’s dollars, Nor-
mates. folk built one for $50 million per mile; Salt Lake City for
After the election, costs leaped upward. In 2009, $56 million per mile; and Sacramento built one for $68
when the FTA approved the projects for PE-entry, RTD million per mile. Minneapolis built one for $105 million
was projecting a total cost of $2.48 billion. RTD’s earli- per mile, which is more typical of recent light-rail projects.
er documents hadn’t predicted first-year ridership but at Yet to be considered by the FTA are Charlotte’s Blue
PE-entry first-year ridership was projected to be 38,600 Line extension, which cost $128 million per mile; the
trips per weekday . Portland-Milwaukie light-rail project, which cost $222
The lines opened in 2013 at a final cost of $2.04 bil- million per mile; and Seattle’s University light-rail proj-
lion. Under the FTA’s current methodology, this would be ect, which cost $628 million per mile as it was all under-
a cost-underrun because it was less than the cost at PE-en- ground. These lines all had small cost overruns and the
try. Yet it cost almost twice what RTD told voters it would Charlotte and Portland projects had large ridership short-
cost in 2004 and what RTD thought it would cost when falls. (Sound Transit, which operates Seattle’s light-rail
it decided to build the lines in 1997. First-year ridership, system, doesn’t report University ridership separately from
incidentally, was fewer than 21,000 people per weekday, the city’s other light-rail line.)
or 46 percent less than what RTD projected at PE-entry. Another source of cost data can be found in the FTA’s
Cost overruns should be calculated by comparing the annual reports on transit capital grants. These reports list
final costs with the projections made at the time the deci- all projects for which transit agencies are seeking or have
sion is made to build the project, which is usually at the received federal grants. Not all projects were built, but
major investment study/analysis of alternatives stage. By they show how much transit agencies thought was reason-
using PE-entry, which is much later, the FTA is signifi- able to spend on rail construction each year.
 Predicted and Actual Costs and Ridership for Selected Federally Funded Transit Projects Built Between 1986 and 2015
Urban Year Route Predicted Actual Differ- Predicted Actual Cost/Mile Inflation
Area Mode Line Completed Miles Cost Cost ence Riders Riders Difference Nominal Adjusted
Washington HR Red & Blue 1986 60.5 4,352 7,968 83% 959,000 762,013 -21% 132 269
Atlanta HR Initial 1987 26.8 1,723 2,720 58% 472,860 222,372 -53% 101 202
Baltimore HR Subway 1987 7.6 804 1,289 60% 103,000 43,044 -58% 170 338
Miami AG Metromover 1988 21.0 84 175 108% 41,000 16,836 -59% 8 16
Detroit AG People Mover 1988 2.9 144 215 49% 67,700 5,928 -91% 74 143
Miami HR Metrorail 1988 21.0 1,008 1,341 33% 239,000 57,530 -76% 64 123
Portland LR Eastside 1988 15.1 172 266 55% 42,500 32,146 -24% 18 34
Sacramento LR Initial 1988 18.3 165 188 14% 50,000 30,326 -39% 10 20
Buffalo LR Metro 1989 6.4 478 722 51% 9,200 19,398 111% 113 209
Pittsburgh LR Reconstruction 1989 10.5 699 622 -11% 90,500 25,733 -72% 59 110
San Diego LR El Cajon 1989 11.1 114 103 -10% 21,600 24,950 16% 9 17
Seattle TB DT Tunnel 1990 1.3 300 469 56% 361 643
San Jose LR Guadalupe 1991 20.0 258 380 48% 41,200 21,035 -49% 19 33
Houston BR Southwest 1993 9.7 96 98 3% 27,280 8,875 -67% 10 17
Chicago HR Southwest 1993 9.0 581 502 -14% 118,760 54,986 -54% 56 92
St. Louis LR Initial 1993 18.0 317 387 22% 41,800 42,381 1% 22 35
Denver BR North I-25 1994 5.3 190 228 20% 43 69
Miami AG Extension 1995 2.5 221 228 3% 20,404 4,158 -80% 91 144
Baltimore HR Hopkins 1995 1.5 314 353 13% 13,600 10,128 -26% 235 372
San Francisco HR Colma 1996 0.9 113 180 60% 15,200 13,060 -14% 197 306
Dallas LR S. Oak Cliff 1996 9.6 325 360 11% 34,170 26,884 -21% 38 58
Baltimore LR BWI HV ext. 1997 7.3 82 116 42% 12,230 8,272 -32% 16 24
San Jose LR Tasman West 1997 7.6 451 325 -28% 14,875 8,244 -45% 43 65
Portland LR Westside 1998 17.7 454 782 72% 60,314 43,876 -27% 44 67
Salt Lake LR I-15 1999 15.0 206 299 45% 26,500 22,100 -17% 20 30
Jacksonville AG Skyway 2000 2.5 66 106 60% 42,472 2,627 -94% 42 62
Pittsburgh BR Airport 2000 6.1 274 322 17% 53 77
Atlanta HR North 2000 3.1 440 473 8% 57,120 20,878 -63% 152 222
Denver LR Southwest 2000 8.7 149 178 19% 22,000 19,083 -13% 20 30
St. Louis LR St. Clair 2001 17.4 368 339 -8% 20,274 15,976 -21% 19 28
Los Angeles HR Red 2002 17.0 3,031 4,470 47% 297,733 134,555 -55% 263 369
Dallas LR North Central 2002 12.5 333 437 31% 17,033 16,278 -4% 35 49
San Francisco HR SFO 2003 8.7 1,283 1,552 21% 67,400 35,534 -47% 178 246
San Francisco HR Airport 2003 8.7 1,194 1,552 30% 68,600 28,321 -59% 178 245
Sacramento LR South 2003 6.3 202 219 8% 12,550 10,543 -16% 35 48
Salt Lake LR University 2003 4.0 189 192 2% 10,050 21,811 117% 48 66
Boston BR Piers 2004 1.0 398 600 51% 24,300 13,298 -45% 600 804
Washington HR Largo 2004 3.1 375 426 14% 14,270 8,623 -40% 138 184
Minneapolis LR Hiawatha 2004 12.0 244 697 186% 37,000 33,477 -10% 58 78
Pittburgh LR Recon 2004 5.5 401 385 -4% 49,000 25,733 -47% 70 94
Portland LR Interstate 2004 5.8 283 350 24% 13,900 11,800 -15% 60 81
Memphis SC Extension 2004 2.0 36 58 61% 4,200 707 -83% 29 39
Chicago HR Douglas recon 2005 6.6 442 441 0% 33,000 28,624 -13% 67 87
San Juan HR Tren Urbano 2005 10.6 1,086 2,228 105% 114,492 31,749 -72% 210 273
San Diego LR Mission Valley 2005 5.9 387 506 31% 10,795 8,895 -18% 86 112
Chicago CR UP West 2006 8.5 99 106 7% 12 16
Chicago CR North Central 2006 55.1 205 217 6% 4 5
Chicago CR Southwest 2006 11.0 179 185 4% 17 21
Baltimore LR Double tracking 2006 9.4 151 152 1% 44,000 28,541 -35% 16 20
Denver LR Southeast 2006 19.1 585 851 45% 38,100 31,320 -18% 44 56
Newark LR Elizabeth I 2006 1.0 181 208 15% 12,500 2,500 -80% 208 262
New Jersey LR Hudson-Bergen 2006 15.4 930 1,756 89% 66,160 41,525 -37% 114 144
Miami CR Double tracking 2007 71.7 330 346 5% 42,100 15,138 -64% 5 6
Charlotte LR Lynx 2007 9.6 331 463 40% 9,100 11,678 28% 48 59
Cleveland BR Euclid 2008 9.4 179 197 10% 21,100 14,300 -32% 21 25
Salt Lake CR Weber 2008 44.0 408 614 50% 8,400 5,300 -37% 14 17
Phoenix LR East Valley 2008 19.7 1,076 1,405 31% 26,000 34,800 34% 71 86
Portland YR WES 2008 14.7 85 162 91% 2,400 1,200 -50% 11 13
San Diego YR Sprinter 2008 22.0 214 478 124% 11,995 6,600 -45% 22 26
Minneapolis CR Northstar 2009 40.0 265 309 16% 4,100 2,200 -46% 8 9
Los Angeles LR Gold line extension 2009 6.0 760 899 18% 150 179
Seattle LR Link 2009 15.6 1,858 2,558 38% 34,900 23,400 -33% 164 196
Dallas LR NW-SE 2010 20.9 1,151 1,406 22% 40,300 32,949 -18% 67 80
Austin BR MetroRapid 2014 34.5 47 39 -17% 11,500 1 1
Flagstaff BR MountainLink 2011 3.4 10 8 -21% 2 3
Dallas LR Northwest 2010 20.9 1,151 1,406 22% 40,300 31,000 -23% 67 80
Portland LR Green Line 2009 8.3 505 576 14% 30,400 24,000 -21% 69 83
Norfolk LR Tide 2011 7.3 195 315 62% 2,900 4,600 59% 43 50
Portland SR Loop 2012 3.3 152 149 -2% 8,100 2,500 -69% 45 51
Phoenix LR Mesa Extension 2015 3.1 199 197 -1% 8,700 8,100 -7% 63 69
Pittsburgh LR North Shore 2012 1.2 327 510 56% 14,300 11,100 -22% 425 483
Salt Lake LR Mid-Jordan 2011 10.6 522 510 -2% 6,300 7,400 17% 48 56
Orlando CR Central Florida 2014 32.0 362 357 -1% 4,300 3,250 -24% 11 12
Sacramento LR South Sacto 2015 4.3 153 270 76% 7,400 4,300 -42% 63 68
Minneapolis LR Central Corridor 2014 9.7 932 927 -1% 32,400 40,400 25% 96 105
Costs are in millions of dollars; ridership is average weekday in the first year after opening, which was not predicted for some projects.
 New Starts Average Light-Rail Costs transit system built in the United States, although Seattle
is on track to beat that record if it ever completes all the
250
Inflation-Adjusted light-rail lines it has on its drawing tables.
The Honolulu rail project is costing far more per mile
Millions of Dollars Per Mile

200
Nominal than any other above-ground rail line built in the United
States. Though grade separated and therefore classified as
150
heavy rail, HART selected a railcar technology with lim-
ited capacity. Given the short platforms used at every sta-
100
tion, it will be able to move no more people than a light-
rail line.
50
A bus line could have moved far more people per hour
for far less cost. Honolulu had originally proposed to build
0
2000 2005 2010 2015 2020 a 32-mile bus-rapid transit line that was projected to cost
Light-rail projects included in the FTA’s annual transit capital grants less than $650 million, or about the cost of one mile of the
(New Starts) reports have tripled in cost in the last two decades. rail line that is now under construction.
I tallied the rail miles and projected construction costs Part of Honolulu’s problem, a state audit revealed,
of all new light-rail construction projects in every report is that HART farmed out 16 senior management posi-
from 2000 to 2022. I limited my review to light rail be- tions to a consulting firm, HDR, paying HDR more than
cause other rail projects can be much more variable. I also $500,000 per manager. The managers then signed hun-
left out projects such as ones in Tacoma and Memphis dreds of change orders, adding half a billion dollars to the
light rail that were called light rail but were really street- project costs but fattening HDR’s revenues.
cars. Since the data for any given year is based on informa- Yet this only explains part of the problem. Another
tion from two years before, I adjusted for inflation using part is that transit planners are guilty of optimism bias,
gross domestic product price deflators from two years be- meaning they tend to make assumptions that favor con-
fore the date of each report. struction rather than no action. “We didn’t lie,” said one of
After adjusting for inflation, the average light-rail cost the planners of the Washington DC Metro, which ended
per mile has tripled since 2000. In 2000, only seven out up costing four times the original projections. “We just
of 20 light-rail proposals cost more than $100 million a used the most optimistic of forecasts.”
mile while nine cost less than $50 million a mile. By 2022, Some planners compound this bias with strategic mis-
none cost less than $100 million a mile and more than half representation, meaning they knowingly lie to the public to
cost more than $200 million a mile. This cost-inflation ap- sell their plans. “I have no apologies to make for overesti-
pears to be the result of transit agencies taking advantage mating ridership and revenue,” said another Washington
of the FTA’s willingness to hand out federal funds for rail Metro planner. “It was in the public interest.”
transit regardless of the cost or cost-effectiveness. A final problem is a sort of Peter Principle of transit:
people who run a halfway-decent bus system—and Ho-
The Honolulu Debacle nolulu’s was one of best bus systems in the country—rise
The tsunami of all cost overruns is in Hawaii, where a 20- to their level of incompetence when they try to plan and
mile rail line in Honolulu was originally projected to cost build a rail system. Rail systems are far more complicated.
less than $3 billion. By 2009, when the FTA agreed to Bus routes can be changed overnight in response to chang-
fund preliminary engineering, the projected cost had risen es in traffic patterns and buses are regularly replaced with
to $5.5 billion and the line was expected to be completed ones using newer technologies. In contrast, rail lines take
in 2019. years to plan and build and railcars have longer lifespans
Today, the cost has risen to $12.4 billion and com- than buses. This means both rail routes and rail technolo-
pletion is not expected until 2031. Making matters worse, gies are likely to be obsolete before they are done.
the Honolulu Authority for Rapid Transit (HART), which Transit agencies try to fix this and create a market for
is building the line, just reduced its ridership projections their billion-dollar white elephants by spending hundreds
by 18 percent based on the decline in Honolulu bus rider- of millions more subsidizing high-density, transit-orient-
ship between 2015 and 2019. No one knows for sure the ed developments. But this has never worked. Portland’s
long-term effects of the pandemic, but it will likely reduce bus system in 1980 carried 10 percent of commuters to
ridership still further. work; by 2019, after spending roughly $5 billion on rail
Urban Honolulu had 834,000 residents in 2019, transit and more than a billion dollars subsidizing tran-
which means the line is costing about $15,000 per resi- sit-oriented developments, transit carried only 8 percent
dent. This is by far the highest cost per capita of any rail of commuters to work. Transit’s share of commuting and/
transit line ever built in the United States. In fact, it is or per capita transit trips similarly declined after Atlanta,
probably less than the capital cost per capita of any rail Baltimore, Dallas, Los Angeles, San Jose, and St. Louis,
among other urban areas, built rail transit and transit-ori-
ented developments. will work with the politicians to keep the money flowing.

Fixing the Problem

Bent Flyvbjerg, a Danish transportation planner who is
now at Oxford University, thinks the solution is reference
class forecasting. This means that, if light rail projects cost
an average of 50 percent more than originally projected,
then all future initial projections should be increased by
50 percent to compensate.
This assumes, however, that people truly understand
big numbers like millions and billions. In fact, any large
number is understood only as an abstraction. The Hono-
lulu rail line was a bad idea when its projected cost was
$3 billion. Yet anyone who nevertheless thought it was a
good idea when it was projected to cost $5 billion prob- One reason average light-rail costs have increased is that Seattle is build-
ing light-rail lines that are almost entirely elevated or underground,
ably wouldn’t have thought any different if the original
including this one under construction in Bellevue. Like the Honolulu
projection was $7.5 billion. line, these lines have the high-cost disadvantage of heavy rail and the
Another idea is to enact firm financial criteria in the low-capacity disadvantage of light rail. Photo by SounderBruce.
federal law authorizing the FTA to fund rail projects. But
such criteria are already there: the 1991 law that autho- The only certain check on cost overruns and oth-
rized such funding specified that grants should be awarded er strategic misrepresentations is to end the subsidies. If
only to transit agencies that had determined that rail tran- transit agencies go broke and transit officials are disgraced
sit was cost effective. This provision was either completely instead of celebrated when cost overruns make projects
ignored or applied only in an extremely weak form that unviable, they will be more careful to curb optimism bias
most transit agencies successfully evaded. The Obama ad- and to ignore strategic misrepresentation. If transit proj-
ministration essentially eliminated the cost effectiveness ects can only be built if there are transit revenues to pay
criteria in a rule approved in 2010. for them, transit agencies will tend to build only the ones
Any criteria written into laws or rules will not with- that truly make sense.
stand certain unfailing political laws: government agencies Randal O’Toole, the Antiplanner, is a transportation and
seek to maximize their budgets; special interest groups land-use policy analyst and author of Romance of the Rails:
seek to get funds from taxpayers; politicians seek campaign Why the Passenger Trains We Love Are Not the Transpor-
contributions to get reelected. So long as there are subsi- tation We Need. Masthead photo of Honolulu rail line un-
dies to be handed out, bureaucrats and special interests der construction is by Musashi1600.