Five Takeaways from Eno’s Transit Capital Construction Database
A growing cadre of researchers, advocates, and experts are asking: why does it cost so much to build infrastructure, particularly public transit, in the United States? Last year, Eno kicked off a major research initiative analyzing cost and timeline drivers affecting the delivery of rail transit in the United States. A crucial aim of this research is to understand if and how costs are high, particularly when compared to other countries.
To learn more about Eno’s transit project delivery research, click here.
To investigate the relative success and challenge of building transit, the Eno team created a database of projects. The database includes construction cost and timeline data for a total of 171 domestic and international rail transit projects completed over the past 20 years. For each project, factors such as number of stations, grade alignment, station spacing, and mode allow for deeper comparisons.
The purpose of this database is to help draw conclusions about the extent to which transit construction costs differ in the United States and peer countries, as well as shed light on the differences between project characteristics and complexity across countries. The initial insights from this data help form the questions and themes we investigate in further detail through regional case studies. The database will be continually updated with more information and insights as they become available.
Geography. The database is limited to examples in the United States, Canada, and Europe. The research team kept the geographical range to these countries and regions because of their comparable political culture, government structures, and infrastructure development and age. Future iterations of this database could include examples beyond these regions, and Eno intends to draw lessons from peer countries in Asia, Oceania, and South America in future research.
Timeframe. The database is largely limited to projects that have been completed between 2000 and 2020. There are some exceptions made on a case-by-case basis to include projects outside this range. For example, the United States and France opened some tunneled transit projects in the 1990s that help provide additional context and comparisons. Similarly, the database generally does not include projects that have not yet opened for service. In part due to common cost overruns, often the full cost of a project is not known until it is completed. But the database does include a few projects in Boston and San Francisco that are set to open in 2021 because of their complexity and importance in the national discussion.
Modes. Defining “modes” of transit is perennial debate, with inconsistencies across and within countries around the world. For the most part, the database focuses on heavy rail and light rail transit projects. Most new transit infrastructure in the United States is light rail, so we included many international examples of light rail projects. In most cases, European trams are similar to U.S. light rail in their grade alignment and right of way.
The database does not include intercity rail projects (like California High Speed Rail or comparable international examples). The database also avoids streetcar projects, which rarely travel in their own right of way and are often loops instead of bidirectional track, making cost comparisons difficult. Some commuter and regional rail projects were included, particularly if they involved building new infrastructure (and are thus similar to heavy rail). But many U.S. commuter rail projects, which primarily run from outlying suburbs to city cores, were also excluded from the database, as the majority of these projects were conversions of existing freight rail infrastructure for commuter rail service, and include little new construction.
Sources. A range of academic, media, industry, and government resources were used to obtain reported construction costs for all new lines entered into the database. It draws from official cost reports wherever possible, either from agencies or other entities directly responsible for construction. When using media reports, we aimed to confirm whether the same – or very similar – cost figure was used across other outlets. Additional project detail collected includes years of groundbreaking and opening for service to the public, project length (kilometers), number of stations, grade alignment (i.e. the share of total alignment that is below ground, at-grade, and above-ground), and station spacing (calculated as average kilometers between stations).
This database also uses inputs from construction cost data collection from the Federal Transit Administration’s Capital Cost Database and by researchers Alon Levy and Eric Goldwyn at the NYU Marron Institute and Yonah Freemark via The Transport Politic.
Adjusting for inflation and conversion to U.S. dollars
To compare projects across geographies and over time, the database adjusts costs so all project costs are compared in 2019 U.S. dollars. This is done with a two-step process.
First, international reported costs were adjusted using purchasing power parity (PPP) rates for projects reported in non-U.S. currency. Currency conversions were based on the OECD’s PPP table, which documents conversion rates for international currencies to U.S. dollars in a given year, taking differing price levels between countries into account (measured as foreign currency needed to purchase $1 worth of goods).
Then, projects were adjusted to 2019 dollars for inflation using the project’s midpoint. Instead of using a standard inflation calculator based on the consumer price index (CPI), the research team decided to use the Bureau of Economic Analysis price index for state and local government investments in transportation, which is a producer price index (PPI). The BEA transportation price index is a more accurate reflection of buying power for investments in infrastructure, as opposed to the CPI, which is based primarily on consumer spending in categories like healthcare, housing, and utilities. (Note that the BEA publishes other producer price indexes for federal transportation and highway costs, which all follow very similar trendlines.)
Figure 1 shows the price indexes for transportation investments and consumer goods grew at a somewhat similar rate from 1980 until 2004, when the cost of transportation investments began to outstrip the increase in the cost of consumer goods. This coincided with a global increase in the cost of key construction materials, such as steel, in the mid-2000s.
Figure 1: CPI vs State/Local Price Index for Fixed Investments in Transportation
One of the most noticeable effects of using the producer price index for transportation instead of CPI is that it significantly boosts the cost (in 2019 dollars) of transit projects completed prior to the mid 2000s. The table below shows how unit costs changed using the different indexes.
Table 1: Changes in Project Unit Cost by Price Index Used
|Project||Unit Cost (millions, 2019 USD) using CPI||Unit Cost (millions, 2019 USD) using PPI for transportation|
|Los Angeles Red Line Phase 1 (1993)||$550/km||$800/km|
|Washington Red Line Extension (1998)||$159/km||$247/km|
|Los Angeles Red line Phase 2A/2B (1999)||$274/km||$449/km|
|Los Angeles Red Line Phase 3 (2000)||$190/km||$277/km|
It is unclear whether the significant increases in project costs in the mid-2000s was the result of materials costs and other factors, or whether project cost increases broadly drove up the price indexes. But the cost increases do seem to be systematic across all transportation modes and perhaps globally.
Adjusting for what costs go into each project
Comparing as-built construction costs can offer some clues as to whether other countries are building public transit systems more cost-effectively. However, there are several caveats and challenges when attempting to make a true “apples to apples” comparison between domestic and international construction costs. The final output of the database is a comparable “unit cost,” in inflation- and currency-adjusted dollars per kilometer of rail line.
But not all projects and agencies are transparent in their cost reporting, and when they are, the data tend to be reported inconsistently. For example, some projects include costs not associated with the actual unit cost of kilometer of rail line. Elements like maintenance facilities or rolling stock are included in some projects, but not others. Worse, detailed cost breakdowns are typically not reported for most projects, and in the event that they are, there may be vast differences in the categories used.
For federally funded projects in the United States, regulations require agencies report cost breakdowns using nine Standard Cost Categories (SCCs):
|10||Guideway & Track Elements|
|20||Stations, Stops, Terminals, Intermodal|
|30||Support Facilities: Yards, Shops, Admin. Bldgs|
|40||Sitework & Special Conditions|
|60||ROW, Land, Existing Improvements|
|Total Project Cost (10-100)|
However, as the Eno team discovered when reviewing select cost breakdowns received through Freedom of Information Act (FOIA) requests, some agencies in the United States also use their own internal methodology to track costs, especially for projects that are locally funded. Rather than reporting project costs for items like stations, sitework, and stations, costs in some cases are broken down by project phase (i.e. preliminary engineering or final design). Cost breakdown methodologies between countries can also vary.
Of the 26 projects in the database that have full cost breakdowns (all U.S. projects), 22 reported vehicles as part of the total cost, and 14 reported some kind of maintenance or support facility. Land acquisition costs were reported in all 26 of the projects, indicating that these are likely included in most U.S. projects. Finding similar cost breakdowns for the rest of the United States and international projects will be part of the next phase of this database. The database does exclude maintenance facilities and rolling stock from total costs when the data is available.
Initial analysis and insights
The caveats and challenges in data reporting outlined above limit the extent to which projects can be thoroughly compared with one another. When comparing construction costs, it is important to avoid drawing sweeping conclusions or over-interpreting trends, and such comparisons will become richer with more data. Keeping these caveats in mind, the following takeaways will inform our research and spark additional questions that in-depth case studies can answer with more accuracy.
Takeaway 1: Light rail is not necessarily cheaper than heavy rail. Grade alignment, rather than mode, is the major determinant of cost.
Defining the mode of a transit project – whether it’s light rail or heavy rail – does not correlate well with its construction cost. Most of the construction and planning inputs for both modes are the same. A transit line, whether heavy or light, includes laying track, installing electrical systems, and building accessible stations. The main difference between the two modes is that light rail tends to be mostly at-grade, and heavy rail is often either tunneled or elevated. However, heavy rail projects outside of the U.S. appear to be largely below grade compared to U.S. projects, as illustrated below.
Figure 2: Grade Alignment Comparision – U.S. and Non-U.S.
The average proportion of heavy rail projects that are at-grade in the United States is over three times higher than non-U.S. heavy rail projects (23 percent vs eight percent, respectively). There is far more similarity among domestic and international light rail/tram projects in the database, though domestic light rail projects appear to include below or above-grade segments more often than international projects in the database. Therefore, when making cost comparisons, light rail is not inherently cheaper than heavy rail – it is only that light rail tends to be at-grade, when heavy rail is usually not, making the latter more expensive.
Takeaway 2: Many rail projects in the United States are relatively inexpensive
High-profile, high-cost projects receive significant attention. The database shows New York City’s Second Ave Subway and 7 line extension cost $2.1 billion per kilometer and $1.8 billion per kilometer, respectively. But for the most part, transit projects in the United States are much less expensive. When excluding the New York outliers, costs per kilometer in the United States average $107 million ($162 million including New York), while non-U.S. projects in the database average $138 million per kilometer. Of the 67 U.S. projects in the database, 43 cost less than $100 million per kilometer, although most of them are built at-grade. These costs are mostly in line with construction costs for similar European and Canadian tram and light rail lines that also run at-grade.
Figure 3 below plots project grade alignments (percent of total alignment that is at-grade) against costs-per-kilometer and illustrates how most U.S. rail transit projects in the database are built primarily at-grade in contrast to non-U.S. projects. This suggests that the United States is able to build fully at-grade transit at a slightly lower average cost than in peer international cities, as discussed further below.
Figure 3: Grade Alignment (Percent At-Grade) vs. Cost-per-Kilometer
Takeaway 3: The United States pays a premium for tunneled projects
Despite some successes domestically and some costly projects abroad, the United States in general pays a significant premium to tunnel, a dynamic that has also caught the attention of some trade publications. Many international projects are below-grade but have similar costs as at-grade projects in the United States.
Projects that are primarily at or above-ground (less than 20 percent of alignment below-ground) cost an average of $73 million per kilometer, which is slightly higher but overall comparable to the average cost of $52 million per kilometer for similar projects abroad. However, projects that are more than 80 percent below ground are built at an average cost of $354 million per kilometer in the United States ($756 million per kilometer including New York City), a 65 percent increase over similar projects abroad ($215 million per kilometer, on average).
The tunneling premium can be seen more clearly in figure 4 below by plotting projects’ share of below-ground alignment with their cost-per-kilometer, and excluding the two outlier projects in New York City from view.* Not only is the cost trendline for U.S. projects much steeper than for non-U.S. projects, but there is a sizeable number of fully tunneled international projects that were built at a comparable cost to U.S. projects in the $100-$200 million per kilometer range.
Figure 4: Percent Tunneled vs. Cost-per-Kilometer
*The New York City projects were not excluded from the plot entirely, but rather excluded from view by restricting the cost ceiling on the Y-axis to allow for a clearer view of the trendline and individual projects below the $500m/km threshold. The projects are still accounted for in the trendlines.
Takeaway 4: Cost variability increases significantly for tunneled projects
Figure 5: Cost Variability by Share of Alignment in Tunnels
Tunneling in particular increases the complexity of a transit project, resulting in much more variability in costs. Figure 5 illustrates the distribution of construction costs-per-kilometer by the share of project alignment below ground. There is noticeable, but not dramatic, variation in construction costs for mostly above-ground projects (<20 percent tunneled) in both the United States and abroad. However, costs can vary considerably for projects that are largely below ground (>80 percent tunneled).
Outside of the United States, where tunneled projects are more common, below-grade lines range from as low as $80-120 million per kilometer for fully underground tram and metro lines in Spain and France, to as high as $300-550 million for subway projects in Barcelona and London (and some Parisian Metro lines).
However, tunneled projects in the United States range from $250-580 million per kilometer (and up to $2 billion for projects in New York City). There are significantly fewer U.S. tunneled lines in the database compared to international projects, and the presence of two large outlier projects in New York City further contributes to the dramatic variation in U.S. costs for tunneled projects. However, current budgets and cost estimates for tunneled lines that are not in the database but are under construction or proposed are still significantly higher than most peer projects abroad, with a notable exception in Seattle.
- Seattle Light Rail Northgate Extension (6.9 km, 5.5 km in tunnels): $273 per km*
- Los Angeles Purple Line Extension Phase 1 (6.4 km): $507 million per km (excl. vehicles)
- Los Angeles Purple Line Extension Phase 2 (4.1 km): $606 million per km (excl. vehicles)
- Los Angeles Purple Line Extension Phase 3 (4.1 km): $870 million per km (excl. vehicles)
- Los Angeles Regional Connector (3.22 km): $559 million per km (excl. vehicles)
- Downtown Austin Light Rail Tunnel (2.4 km): $833 million per kilometer
*Cost obtained via FOIA request
If included in the database, these projects would still fall within the higher cost-range shown in the boxplot above for U.S. projects, and add further examples of the relatively high cost of building below-ground transit in the United States.
Some of the cost variation for tunneled projects can be attributed to factors like geological conditions (which vary considerably in each region and can significantly influence the cost and complexity of tunnel boring), technical specifications, tunnel depth, or station design. The Eno team’s detailed, regional case studies will shed light on other governance or process-related elements that can affect construction costs, including project and contractor management, institutional expertise, permitting, and regulation.
Takeaway 5: Stations are expensive, but international projects include more of them
Stations can also constitute a large portion of overall transit project costs. For tunneled projects in the United States, the database shows stations accounting for around 25 percent of total project costs. Research shows that station depth and architecture is a potential project cost driver. But despite their generally lower cost per kilometer, international projects have closer station spacing on average, which is usually more common and useful in dense downtown areas. However, the database analysis shows station spacing does not seem to have a clear correlation with cost.
Figure 6: Average Distance (km) between Stations – U.S. vs. Non-U.S.
The database calculates the average distance*, in kilometers, between stations. A high-level comparison of station spacing across U.S. and non-U.S. project above suggests that transit stations are spaced closer together abroad, especially for lines mostly at-grade, which have nearly a third of the distance between stations as at-grade U.S. lines. These at-grade lines – most of which are tram or light rail projects – often run through dense, historic city centers and are usually not grade-separated.
*Note: for extensions of existing lines, the total length of a project’s alignment was divided by the total number of stations. For new lines, the total length of the alignment was divided by the number of stations minus one to account for terminal stations at the end of the line.
Figure 7: Station Spacing vs. Cost-per-Kilometer
Comparing average station spacing of projects with their cost-per-kilometer does not indicate a relationship between station spacing and costs, but suggests that European transit projects have higher station densities without a significant cost premium. Most U.S. light rail lines appear to have one to three kilometers between stations, while most non-U.S. light rail/tram lines have 0.4 to 1 kilometer between stations. This comparison, however, may not fully capture differences in technical complexity between U.S. and non-U.S. projects, particularly considering that some international tram lines might have more in common with mixed-traffic streetcars compared to fully grade-separated light rail in the United States.
On the other hand, there appears to be wide variation in station spacing among non-U.S. heavy rail projects – ranging from an average of 0.5 to 2 kilometers between stations – while most U.S. heavy rail lines appear to have an average of 2 to 5 kilometers between stations. This suggest that other countries are building below-grade at a comparable cost to some at-grade U.S. projects while also having higher station densities.
The in-depth case studies currently being conducted by the Eno team will help us better understand the extent to which there is a tunneling premium in the United States, and whether technical standards, governance, project management, contracting practices, regulation, or other factors may explain the large share of relatively inexpensive tunneled rail lines outside of the U.S.
The Eno team will continue to collect additional data and documents on project costs, complexity, and technical standards to make a more accurate comparison between projects. We also intend to continue adding additional projects to the database, particularly more light rail projects in Western Europe, Canada, and Australia. We will announce new updates to the database and any new work products in the future.