From San Francisco’s original BART line in the 1970s to the recently opened Tucson streetcar, rail transit projects have consistently cost an average of about 50 percent more than the projections at the time the decision was made to build the lines. Not all rail projects have gone over budget, but others have cost more than twice the original projections. In contrast, a review of American highway projects found that they tend to go an average of about 8 percent over their original projections.
Rail project planning generally takes the following steps:
- Development and analysis of a wide range of alternatives, usually including bus-rapid transit and sometimes even highway improvements. This used to be called the “major investment study” but more recently has been called the “alternatives analysis.”
- Based on the alternatives analysis, all alternatives are discarded but rail, a “no-build” alternative, and something called “transportation system management” (TSM) which basically means slight upgrades to the existing bus system. The reasons for selecting rail as the “locally preferred alternative” are usually fairly specious: “yes, rail costs ten times as much as bus improvements, but it is projected to attract 1 percent more transit riders.”
- After eliminating most of the alternatives, the next step is to do preliminary engineering for the rail project. This usually identifies a huge increase in costs. Planners may also change the project by, for example, adding new stations to the transit line or deciding to use a more expensive method of construction or propulsion. Despite the higher costs, no alternatives other than no-build and TSM are usually considered.
- Planners then write the draft and final environmental impact statements. This is generally followed by final engineering studies that pinpoint the real cost.
Transit agencies often claim that they completed their rail projects under budget. What they mean is that the project cost less than the final engineering estimates, not the original projections. Since the decision to build rail is usually made based on those original projections, they, not the later estimates, should be the standard.
Although the problem of rail cost overruns was first pointed out in a 1990 report released by the Department of Transportation, it hasn’t been fixed. Projects in the 2000s have gone just as much over their original projections as projects in the 1980s.
For example, when Denver’s Regional Transit District (RTD) decided to build the West light-rail line, it projected that it would cost about $250 million. After adjusting for inflation, that’s about $350 million in 2012 dollars, the year the line was completed for $707 million. Moreover, to build the line for $707 million, RTD ended up putting in single track for the last several miles of the line instead of the double track originally planned.
|Urban Area||Mode||Route||Year Open||Predicted Cost||Actual Cost||Overrun|
|Washington||HR||Red & Blue||1986||4,352||7,968||83%|
|San Diego||LR||El Cajon||1989||114||103||-10%|
|Dallas||LR||S Oak Cliff||1996||325||360||11%|
|Baltimore||LR||BWI HV ext.||1997||82||116||42%|
|San Jose||LR||Tasman West||1997||281||325||16%|
|St. Louis||LR||St. Clair||2001||256||339||32%|
|San Diego||LR||Mission Vly||2005||387||506||31%|
|San Juan||HR||Tren Urbano||2005||1,086||2,228||105%|
|New Jersey||LR||H-B 1 & 2||2006||930||1,756||89%|
|Los Angeles||LR||Gold ext.||2009||760||899||18%|
Most of the numbers in the above table are from a series of Department of Transportation reports, the first of which came out in 1990 (23.5 MB), with updates in 2003 (17.2 MB), 2007 (3.7 MB), 2008 (0.3 MB), 2011 (0.7 MB), 2012 (0.1 MB), and 2013 (0.8 MB).
The above table includes five projects that weren’t in the DOT reports. One, the Los Angeles Blue Line, is documented in a 2006 paper by researchers at Northeastern University. All of the numbers in the paper are identical to the numbers in the 2007 DOT report except that the paper also has the Blue Line while the DOT report does not. The methodologies should be identical and the rail projects in the DOT reports and the Northeastern University paper are a representative sample, as the FTA would be unlikely to deliberately bias the sample to projects that went over cost by more than the average.
The other four projects were completed since 2009, the date of the last project studied in the DOT papers: Austin, Denver West, Norfolk, and Minneapolis Green. The numbers in the table are based on FTA New Starts reports and other official documentation.
All of these numbers compare the estimated cost based on preliminary engineering at the time the agency proposing the line compared it with various alternatives. For older projects, that stage was called the major investment study; for newer ones, it was called the alternatives analysis. The list only includes new construction. Reconstruction projects and double-tracking of existing lines mostly come in pretty close to the original projected budgets.
The projected costs for two projects were adjusted based on major changes to the proposed project. The San Jose Tasman West and St. Louis St. Clair projects were reported by DOT to cost significantly less than the original projected cost, but both of these projects were significantly reduced in length between the original projection and the final project. To account for this, the cost shown in the table above is the projected cost per mile times the number of miles actually built. Tasman West was projected to cost $451 million for 12.2 miles; so the projected cost for the 7.6 miles would have been about $281 million. St. Clair was projected to cost $368 million for 25 miles, but only 17.4 was built which would have cost about $256 million.
The table shows that cost projections have not significantly improved over the last 30 years. While the average overrun in the 2010s may be a bit high due to the projects shown not being a representative sample, it doesn’t seem likely that adding more would greatly improve the results. Four bus projects are included in the table; two of them came in less than 20 percent over projected costs, but the number may be too few to be a representative sample.